Data Mining 2019 – 2020
This is an overview of the slides of the courses Data Mining, added with notes off the book. Some notes of the classes are in
grey.
Overall goals for this course
• You know how to solve business problems by data-analytical thinking
• You are able to interact competently on the topic of data science for business analytics
o For example: is there a data science angle to Facebook’s 19 billion US$ acquisition of WhatsApp?
o For example: why did Match Inc. share price suddenly drop after Facebook revealed its dating app?
• You know several tools and ways of how to practically implement solution methods
What will we be seeing in the course?
• Predictive modeling
o Data Preprocessing
o Terminology
o Decision Trees and Support Vector Machines
• Desciptive modeling
o Clustering
o Association Rule mining
o Recommender Systems
• Evaluating models
• Text Mining
• Deep learning and Vision
• Applications in Marketing, Risk Management, HR, and others
• Intro to Data science and Ethics
Overview Data Mining course
• Intro to Data Mining and Applications
• Data Preprocessing
• Predictive Modeling
o Decision Trees
o Learning versus Overfitting
• Evaluating Classifiers
o Accuracy, Confusion Matrix and Expected Profit
o Visualizing Model Performance
• Naive Bayes
• Big Data
• Text Mining Descriptive Modeling
o Clustering and Association Rule Mining
o Recommender Systems
• Support Vector Machines
• Artificial Neural Networks and Deep Learning
• Ensemble Methods
• Ethics of Data Science
1
,Inhoudsopgave Data Mining 2019-2020
LECTURE 1: Introduction and applications (chapters 1-2) ................................................................................................... 6
Chapter 1 book .................................................................................................................................................................. 7
LECTURE 2: to explain vs to predict and preprocessing ..................................................................................................... 11
1. Explaining versus predicting ......................................................................................................................................... 11
2. Data preprocessing ...................................................................................................................................................... 13
2.1 Sampling.......................................................................................................................................................................... 13
Intermezzo: lazy learner 1NN ............................................................................................................................................ 14
2.2 Encoding .......................................................................................................................................................................... 14
2.3 Missing values ................................................................................................................................................................. 14
2.4 Outliers ............................................................................................................................................................................ 15
Intermezzo: lazy learner 1NN ............................................................................................................................................ 15
2.5 Normalising ..................................................................................................................................................................... 15
2.6 Discretisation and grouping of attribute values .............................................................................................................. 16
Chapter 2 book ................................................................................................................................................................ 17
LECTURE 3: decision trees and overfitting (chapters 3-5) .................................................................................................. 20
Chapter 3: introduction to predictive modeling ............................................................................................................... 20
1. Terminology ...................................................................................................................................................................... 20
2. Finding informative variables from the data ..................................................................................................................... 20
3. Decision trees .................................................................................................................................................................... 21
Methodology of decision trees in more detail (not in book) ............................................................................................. 23
1. Splitting rules..................................................................................................................................................................... 23
2. STOPPING RULES ............................................................................................................................................................... 26
3. Assignment rule ................................................................................................................................................................. 27
Decision trees: advantages and disadvantages .................................................................................................................... 27
Chapter 5: Overfitting and its avoidance (not in book) ..................................................................................................... 28
1. Fitting graph ...................................................................................................................................................................... 28
2. Predictive models: attention points................................................................................................................................... 29
2.1 Training set versus Test set ......................................................................................................................................... 29
2.2 Learning versus Overfitting ......................................................................................................................................... 30
2.3 Generalisation behaviour ............................................................................................................................................ 30
2
,LECTURE 4: assessing and visualizing model performance (chapter 7-8) ........................................................................... 32
Chapter 7: decision analytic thinking: what is a good model?........................................................................................... 32
1. Evaluating classifiers ......................................................................................................................................................... 32
1.1 Accuracy ...................................................................................................................................................................... 32
1.2 Confusion matrix ......................................................................................................................................................... 32
1.3 Problems : unbalanced classes .................................................................................................................................... 32
1.4 Problems : unequal costs and benefits ....................................................................................................................... 32
2. Expected value .................................................................................................................................................................. 33
2.1 Expected Value for classifier evaluation ..................................................................................................................... 33
2.2 Exercises ...................................................................................................................................................................... 35
3. Evaluation and baseline performance ............................................................................................................................... 36
Chapter 8: visualizing model performance ....................................................................................................................... 37
1. Introduction ....................................................................................................................................................................... 37
2. Profit curves ...................................................................................................................................................................... 37
3. ROC curve .......................................................................................................................................................................... 38
4. AUC.................................................................................................................................................................................... 38
5. Cumulative response and lift curves .................................................................................................................................. 38
5.1 cumulative response curve ......................................................................................................................................... 38
5.2 lift curve ...................................................................................................................................................................... 38
6. Exercices ............................................................................................................................................................................ 39
LECTURE 5: naïve bayes (chapter 9) ................................................................................................................................. 42
1. Introduction: Evidence ....................................................................................................................................................... 42
2. Bayes rule and naïve bayes ............................................................................................................................................... 42
3. Exercises ............................................................................................................................................................................ 43
LECTURE 6: descriptive data mining (chapters 6, 12) ........................................................................................................ 45
1. Nearest-neighbor ......................................................................................................................................................... 45
2. Clustering .................................................................................................................................................................... 47
3. Frequent itemsets and Association rules (not in book) ................................................................................................. 48
Apriori – Step 1 ...................................................................................................................................................................... 49
Apriori – Step 2 ...................................................................................................................................................................... 49
4. Recommender systems (not in book) ........................................................................................................................... 50
4.1 Recommender systems – utility function Content-based recommendation ................................................................... 50
4.2 Recommender systems – utility function Collaborative recommendation ..................................................................... 51
4.3 Recommender systems – dimensionality reduction ........................................................................................................ 51
4.4 In conclusion .................................................................................................................................................................... 51
5. Exercises ...................................................................................................................................................................... 52
5.1 Apriori.............................................................................................................................................................................. 52
5.2 Clustering – hierarchical .................................................................................................................................................. 53
5.3 Clustering – k-means ....................................................................................................................................................... 53
Some example questions ................................................................................................................................................. 54
3
,LECTURE 7: Text mining (chapter 10)................................................................................................................................ 55
1. Why text mining? .............................................................................................................................................................. 55
2. Text preprocessing ............................................................................................................................................................ 55
3. Document Classification and Clustering ............................................................................................................................ 58
3.1 document classification............................................................................................................................................... 58
3.2 document clustering ................................................................................................................................................... 58
4. Topic Modeling and Word Embeddings ............................................................................................................................ 59
4.1 topic models ................................................................................................................................................................ 59
4.2 word embeddings ....................................................................................................................................................... 60
5. Case Study in Politics ......................................................................................................................................................... 61
6. Exercise HR analytics ......................................................................................................................................................... 63
LECTURE 8: text mining OLS, logit .................................................................................................................................... 64
1. OLS – ordinary least squares regression ............................................................................................................................ 64
2. Ridge regression ................................................................................................................................................................ 64
3. Logistic regression ............................................................................................................................................................. 65
4. Gradient descent ............................................................................................................................................................... 67
Exercises ................................................................................................................................................................................ 68
Questions .............................................................................................................................................................................. 68
LECTURE 9: artifical neural networks and deep learning Ensemble methods .................................................................... 69
1. Ensemble methods ............................................................................................................................................................ 69
Combine by consensus ..................................................................................................................................................... 70
1.1 Bagging ........................................................................................................................................................................ 70
1.2 Random Forests .......................................................................................................................................................... 70
Combine by Learning ........................................................................................................................................................ 70
1.3 Boosting ...................................................................................................................................................................... 70
1.4 Stacking ....................................................................................................................................................................... 71
Conclusion......................................................................................................................................................................... 71
2. Artificial neural networks .................................................................................................................................................. 72
3. Deep learning .................................................................................................................................................................... 76
Questions .............................................................................................................................................................................. 77
LECTURE 10: DEEP LEARNING ........................................................................................................................................... 78
1. DEEP LEARNING................................................................................................................................................................. 78
2. THE SVM – Support vector machines ................................................................................................................................ 79
3. ADVANCED TOPIS .............................................................................................................................................................. 80
3.1 networked data ........................................................................................................................................................... 80
3.2 the black swan............................................................................................................................................................. 82
3.3 future of Al .................................................................................................................................................................. 82
Resources Al ..................................................................................................................................................................... 83
Use Al ................................................................................................................................................................................ 83
4
,LECTURE 11: Data science ethics ...................................................................................................................................... 84
1. Data gathering: privacy, A/B, Testing and bias ................................................................................................................. 85
2. Data processing: proxies, Government backdoors ............................................................................................................ 87
2.1 proxies ......................................................................................................................................................................... 87
2.2 government backdoors ............................................................................................................................................... 87
3. Modeling: ZK proofs, homomorphic encryption, discrimination ....................................................................................... 88
3.1 Zero Knowledge Proof ................................................................................................................................................. 88
3.2 Encryption ................................................................................................................................................................... 88
3.3 discrimination.............................................................................................................................................................. 88
4. Model evaluation: explain ................................................................................................................................................. 89
4.1 global explanations ..................................................................................................................................................... 89
4.2 instance-based explanations ....................................................................................................................................... 90
5. Deployment: unintended consequences ............................................................................................................................ 91
5. 1 Unintended consequences ......................................................................................................................................... 91
5.2 Cautionary tales .......................................................................................................................................................... 92
5.3 Conclusion ................................................................................................................................................................... 92
LECTURE 12: Recap and questions.................................................................................................................................... 93
1. Example Exam Questions .................................................................................................................................................. 93
2. Questions........................................................................................................................................................................... 93
5
, LECTURE 1: INTRODUCTION AND APPLICATIONS (CHAPTERS 1-2)
• Example: Hurricane Frances
o Why would data-driven prediction be useful? To project the amount of increase in sales due to the
hurricane, to ensure that local Walmart’s are properly stocked
o Goal: find non-obvious patterns
• Terminology
o Data science: a set of fundamental principles that guide the extraction of knowledge from data
o Data mining: automatic extraction of patterns from large amounts of data (via tools/technologies that
incorporate the principles)
o Big Data: data that is so large that traditional data processing systems are unable to deal with it (both
storage and analysis component)
o Querying and reporting
§ You know exactly what you are looking for.
§ SQL (commando’s such as select, from, where…)
o OLAP – On-Line Analytical Processing
§ = Advanced query and reporting
§ Multidimensional analysis
Terminology
§ Nice visualization, data cubes, roll-up, slice and dice, ...
§ Resembles with query, but OLAP has also visualization
14
Business Intelligence
§ Can show problems that exist in data
o Getting the right information to the right person at the right
o Business Intelligence = getting the right information to the right person at the right time
time
You know what you Business intelligence includes data
OLAP look for! SQL
warehousing, OLAP and Data
Data ware- mining.
housing
Data warehousing tries to solve the
problem in all large companies,
You don’t know what
you look for! Looking
such as the fact that data is stored
for new knowledge in different databases over many
different locations.
Data mining
“collect and coalesce data from across an enterprise,
often from multiple transaction-processing systems, Data warehousing will collect data
15
each with its own database.”
in 1 warehouse
§ Data mining: Setting the scene with credit scoring example
• Banks: should I grant credit to this loan applicant?
• Predict the creditworthiness, based on historical data (not possible when new bank)
§ Difference data mining and OLAP: an algorithm looks for data in data mining, and in OLAP you
have to define what you are looking for
o Machine Learning
§ Improving the knowledge of a learning agent (computer), which try to find patterns based on
a Mining §
data
More than just data mining, also computer vision and robotics
o Artificial Intelligence (Al)
§ A computer interacts through data
ata mining: automatic extraction of knowledge from data § Learning from data leads to intelligence
§ Big Data + Machine Learning = Artificial Intelligence (Theodoros Evgeniou, 2019)
tting the scene with credit scoring example § Renewed interest from Deep Learning (large artificial neural networks)
Banks: should I grant credit §
to this loanMost work in AI is on data mining
applicant? Notations
o The separation between the fields has blurred
Predit the creditworthiness, based on historical data
Data
Input matrix X Target variable Y
Data mining technique
Data instance xi
a vector of size m (number of input variables)
Pattern i = 1, 2, …, m (number of data instances)
Note: an intial set of data instances with known target 6
variable needed! Input variable or feature xj
16 a vector of size n (number of data instances)
xij: the value for input variable j i = 1, 2, …, m (number of input variables)
for data instance i
This is an overview of the slides of the courses Data Mining, added with notes off the book. Some notes of the classes are in
grey.
Overall goals for this course
• You know how to solve business problems by data-analytical thinking
• You are able to interact competently on the topic of data science for business analytics
o For example: is there a data science angle to Facebook’s 19 billion US$ acquisition of WhatsApp?
o For example: why did Match Inc. share price suddenly drop after Facebook revealed its dating app?
• You know several tools and ways of how to practically implement solution methods
What will we be seeing in the course?
• Predictive modeling
o Data Preprocessing
o Terminology
o Decision Trees and Support Vector Machines
• Desciptive modeling
o Clustering
o Association Rule mining
o Recommender Systems
• Evaluating models
• Text Mining
• Deep learning and Vision
• Applications in Marketing, Risk Management, HR, and others
• Intro to Data science and Ethics
Overview Data Mining course
• Intro to Data Mining and Applications
• Data Preprocessing
• Predictive Modeling
o Decision Trees
o Learning versus Overfitting
• Evaluating Classifiers
o Accuracy, Confusion Matrix and Expected Profit
o Visualizing Model Performance
• Naive Bayes
• Big Data
• Text Mining Descriptive Modeling
o Clustering and Association Rule Mining
o Recommender Systems
• Support Vector Machines
• Artificial Neural Networks and Deep Learning
• Ensemble Methods
• Ethics of Data Science
1
,Inhoudsopgave Data Mining 2019-2020
LECTURE 1: Introduction and applications (chapters 1-2) ................................................................................................... 6
Chapter 1 book .................................................................................................................................................................. 7
LECTURE 2: to explain vs to predict and preprocessing ..................................................................................................... 11
1. Explaining versus predicting ......................................................................................................................................... 11
2. Data preprocessing ...................................................................................................................................................... 13
2.1 Sampling.......................................................................................................................................................................... 13
Intermezzo: lazy learner 1NN ............................................................................................................................................ 14
2.2 Encoding .......................................................................................................................................................................... 14
2.3 Missing values ................................................................................................................................................................. 14
2.4 Outliers ............................................................................................................................................................................ 15
Intermezzo: lazy learner 1NN ............................................................................................................................................ 15
2.5 Normalising ..................................................................................................................................................................... 15
2.6 Discretisation and grouping of attribute values .............................................................................................................. 16
Chapter 2 book ................................................................................................................................................................ 17
LECTURE 3: decision trees and overfitting (chapters 3-5) .................................................................................................. 20
Chapter 3: introduction to predictive modeling ............................................................................................................... 20
1. Terminology ...................................................................................................................................................................... 20
2. Finding informative variables from the data ..................................................................................................................... 20
3. Decision trees .................................................................................................................................................................... 21
Methodology of decision trees in more detail (not in book) ............................................................................................. 23
1. Splitting rules..................................................................................................................................................................... 23
2. STOPPING RULES ............................................................................................................................................................... 26
3. Assignment rule ................................................................................................................................................................. 27
Decision trees: advantages and disadvantages .................................................................................................................... 27
Chapter 5: Overfitting and its avoidance (not in book) ..................................................................................................... 28
1. Fitting graph ...................................................................................................................................................................... 28
2. Predictive models: attention points................................................................................................................................... 29
2.1 Training set versus Test set ......................................................................................................................................... 29
2.2 Learning versus Overfitting ......................................................................................................................................... 30
2.3 Generalisation behaviour ............................................................................................................................................ 30
2
,LECTURE 4: assessing and visualizing model performance (chapter 7-8) ........................................................................... 32
Chapter 7: decision analytic thinking: what is a good model?........................................................................................... 32
1. Evaluating classifiers ......................................................................................................................................................... 32
1.1 Accuracy ...................................................................................................................................................................... 32
1.2 Confusion matrix ......................................................................................................................................................... 32
1.3 Problems : unbalanced classes .................................................................................................................................... 32
1.4 Problems : unequal costs and benefits ....................................................................................................................... 32
2. Expected value .................................................................................................................................................................. 33
2.1 Expected Value for classifier evaluation ..................................................................................................................... 33
2.2 Exercises ...................................................................................................................................................................... 35
3. Evaluation and baseline performance ............................................................................................................................... 36
Chapter 8: visualizing model performance ....................................................................................................................... 37
1. Introduction ....................................................................................................................................................................... 37
2. Profit curves ...................................................................................................................................................................... 37
3. ROC curve .......................................................................................................................................................................... 38
4. AUC.................................................................................................................................................................................... 38
5. Cumulative response and lift curves .................................................................................................................................. 38
5.1 cumulative response curve ......................................................................................................................................... 38
5.2 lift curve ...................................................................................................................................................................... 38
6. Exercices ............................................................................................................................................................................ 39
LECTURE 5: naïve bayes (chapter 9) ................................................................................................................................. 42
1. Introduction: Evidence ....................................................................................................................................................... 42
2. Bayes rule and naïve bayes ............................................................................................................................................... 42
3. Exercises ............................................................................................................................................................................ 43
LECTURE 6: descriptive data mining (chapters 6, 12) ........................................................................................................ 45
1. Nearest-neighbor ......................................................................................................................................................... 45
2. Clustering .................................................................................................................................................................... 47
3. Frequent itemsets and Association rules (not in book) ................................................................................................. 48
Apriori – Step 1 ...................................................................................................................................................................... 49
Apriori – Step 2 ...................................................................................................................................................................... 49
4. Recommender systems (not in book) ........................................................................................................................... 50
4.1 Recommender systems – utility function Content-based recommendation ................................................................... 50
4.2 Recommender systems – utility function Collaborative recommendation ..................................................................... 51
4.3 Recommender systems – dimensionality reduction ........................................................................................................ 51
4.4 In conclusion .................................................................................................................................................................... 51
5. Exercises ...................................................................................................................................................................... 52
5.1 Apriori.............................................................................................................................................................................. 52
5.2 Clustering – hierarchical .................................................................................................................................................. 53
5.3 Clustering – k-means ....................................................................................................................................................... 53
Some example questions ................................................................................................................................................. 54
3
,LECTURE 7: Text mining (chapter 10)................................................................................................................................ 55
1. Why text mining? .............................................................................................................................................................. 55
2. Text preprocessing ............................................................................................................................................................ 55
3. Document Classification and Clustering ............................................................................................................................ 58
3.1 document classification............................................................................................................................................... 58
3.2 document clustering ................................................................................................................................................... 58
4. Topic Modeling and Word Embeddings ............................................................................................................................ 59
4.1 topic models ................................................................................................................................................................ 59
4.2 word embeddings ....................................................................................................................................................... 60
5. Case Study in Politics ......................................................................................................................................................... 61
6. Exercise HR analytics ......................................................................................................................................................... 63
LECTURE 8: text mining OLS, logit .................................................................................................................................... 64
1. OLS – ordinary least squares regression ............................................................................................................................ 64
2. Ridge regression ................................................................................................................................................................ 64
3. Logistic regression ............................................................................................................................................................. 65
4. Gradient descent ............................................................................................................................................................... 67
Exercises ................................................................................................................................................................................ 68
Questions .............................................................................................................................................................................. 68
LECTURE 9: artifical neural networks and deep learning Ensemble methods .................................................................... 69
1. Ensemble methods ............................................................................................................................................................ 69
Combine by consensus ..................................................................................................................................................... 70
1.1 Bagging ........................................................................................................................................................................ 70
1.2 Random Forests .......................................................................................................................................................... 70
Combine by Learning ........................................................................................................................................................ 70
1.3 Boosting ...................................................................................................................................................................... 70
1.4 Stacking ....................................................................................................................................................................... 71
Conclusion......................................................................................................................................................................... 71
2. Artificial neural networks .................................................................................................................................................. 72
3. Deep learning .................................................................................................................................................................... 76
Questions .............................................................................................................................................................................. 77
LECTURE 10: DEEP LEARNING ........................................................................................................................................... 78
1. DEEP LEARNING................................................................................................................................................................. 78
2. THE SVM – Support vector machines ................................................................................................................................ 79
3. ADVANCED TOPIS .............................................................................................................................................................. 80
3.1 networked data ........................................................................................................................................................... 80
3.2 the black swan............................................................................................................................................................. 82
3.3 future of Al .................................................................................................................................................................. 82
Resources Al ..................................................................................................................................................................... 83
Use Al ................................................................................................................................................................................ 83
4
,LECTURE 11: Data science ethics ...................................................................................................................................... 84
1. Data gathering: privacy, A/B, Testing and bias ................................................................................................................. 85
2. Data processing: proxies, Government backdoors ............................................................................................................ 87
2.1 proxies ......................................................................................................................................................................... 87
2.2 government backdoors ............................................................................................................................................... 87
3. Modeling: ZK proofs, homomorphic encryption, discrimination ....................................................................................... 88
3.1 Zero Knowledge Proof ................................................................................................................................................. 88
3.2 Encryption ................................................................................................................................................................... 88
3.3 discrimination.............................................................................................................................................................. 88
4. Model evaluation: explain ................................................................................................................................................. 89
4.1 global explanations ..................................................................................................................................................... 89
4.2 instance-based explanations ....................................................................................................................................... 90
5. Deployment: unintended consequences ............................................................................................................................ 91
5. 1 Unintended consequences ......................................................................................................................................... 91
5.2 Cautionary tales .......................................................................................................................................................... 92
5.3 Conclusion ................................................................................................................................................................... 92
LECTURE 12: Recap and questions.................................................................................................................................... 93
1. Example Exam Questions .................................................................................................................................................. 93
2. Questions........................................................................................................................................................................... 93
5
, LECTURE 1: INTRODUCTION AND APPLICATIONS (CHAPTERS 1-2)
• Example: Hurricane Frances
o Why would data-driven prediction be useful? To project the amount of increase in sales due to the
hurricane, to ensure that local Walmart’s are properly stocked
o Goal: find non-obvious patterns
• Terminology
o Data science: a set of fundamental principles that guide the extraction of knowledge from data
o Data mining: automatic extraction of patterns from large amounts of data (via tools/technologies that
incorporate the principles)
o Big Data: data that is so large that traditional data processing systems are unable to deal with it (both
storage and analysis component)
o Querying and reporting
§ You know exactly what you are looking for.
§ SQL (commando’s such as select, from, where…)
o OLAP – On-Line Analytical Processing
§ = Advanced query and reporting
§ Multidimensional analysis
Terminology
§ Nice visualization, data cubes, roll-up, slice and dice, ...
§ Resembles with query, but OLAP has also visualization
14
Business Intelligence
§ Can show problems that exist in data
o Getting the right information to the right person at the right
o Business Intelligence = getting the right information to the right person at the right time
time
You know what you Business intelligence includes data
OLAP look for! SQL
warehousing, OLAP and Data
Data ware- mining.
housing
Data warehousing tries to solve the
problem in all large companies,
You don’t know what
you look for! Looking
such as the fact that data is stored
for new knowledge in different databases over many
different locations.
Data mining
“collect and coalesce data from across an enterprise,
often from multiple transaction-processing systems, Data warehousing will collect data
15
each with its own database.”
in 1 warehouse
§ Data mining: Setting the scene with credit scoring example
• Banks: should I grant credit to this loan applicant?
• Predict the creditworthiness, based on historical data (not possible when new bank)
§ Difference data mining and OLAP: an algorithm looks for data in data mining, and in OLAP you
have to define what you are looking for
o Machine Learning
§ Improving the knowledge of a learning agent (computer), which try to find patterns based on
a Mining §
data
More than just data mining, also computer vision and robotics
o Artificial Intelligence (Al)
§ A computer interacts through data
ata mining: automatic extraction of knowledge from data § Learning from data leads to intelligence
§ Big Data + Machine Learning = Artificial Intelligence (Theodoros Evgeniou, 2019)
tting the scene with credit scoring example § Renewed interest from Deep Learning (large artificial neural networks)
Banks: should I grant credit §
to this loanMost work in AI is on data mining
applicant? Notations
o The separation between the fields has blurred
Predit the creditworthiness, based on historical data
Data
Input matrix X Target variable Y
Data mining technique
Data instance xi
a vector of size m (number of input variables)
Pattern i = 1, 2, …, m (number of data instances)
Note: an intial set of data instances with known target 6
variable needed! Input variable or feature xj
16 a vector of size n (number of data instances)
xij: the value for input variable j i = 1, 2, …, m (number of input variables)
for data instance i
