Task 2: Part C -the
application
Contents
What does the application need to do?
Coding
Application Performance
FAQ
Questions, comments, or suggestions?
Your software application (the “app”) is the entirety of part C of task 2.
However, the Computer Science capstone is not a software project. Other than
requiring you to apply machine learning (ML) to data, they will only assess
what your application does -not its presentation.
Tip
The official rubric and directions were written to map the project’s
elements to specific competencies. However, to allow for a broad
range of projects its language is necessarily also broad. For more
specifics, we recommend referring to the guidelines on this webpage
and the Task 2 template.
,What does the application need to
do?
Your app must help the user solve the organizational problem from task 1 by
doing the following:
1. Data → ML model: Use data to develop a machine learning model.
2. Accuracy Metric: Provide an appropriate metric or plan for measuring
the app’s accuracy.
3. Visualizations: Present 3 different pictures; images can code generated
or static.
4. User Application: Provide a way for the “user” to use the ML model
towards solving the problem.
Example 1 Example 2 Example 3 Example 4
The App: A standalone Jupyter Notebook file (.ipynb) which can predict Iris
types (see the example).
Data → ML model: Labeled Iris petal dimensions train an SVM
classification model.
Accuracy Metric: Percent of correct predictions on testing data.
Visualizations: Histograms showing distributions of different flower
features and a confusion matrix illustrating the accuracy of the
classification model.
User interface: (New data → ML model → prediction) Following
detailed instructions, the user can input flower dimensions, run code
in the Jupyter notebook, and the app returns a prediction helping
customers identify their flowers.
, Data requirements
Your data must be sufficiently complex and processed for your ML
algorithm to function as needed.
Evaluators must be able to access your data.
There are no specific requirements regarding the complexity or nature of your
data. However, your application must work and fulfill the organizational need
you’ll outline in the task 2 documentation. Therefore, when choosing your raw
data, consider carefully any additional time and effort necessary to prepare
that data for use. As these extra steps are only needed in so far as the chosen
problem and algorithm need them, you may wish to adjust your project or
choose a different data set accordingly.
You do not need special permission to use any open-source dataset (see our
list). Furthermore, data sets used in previous C964 projects are available for
reuse (no list of previously used datasets exists). It’s only required that the
data meet the needs of your project and be legally available to use and share
with evaluators. For example, using data belonging to a current employer
would require submitting a waiver form.
You can use data used for previous projects (submitted by you or others)
including any dataset found on kaggle.com.
You do not have to use all the data. Removing columns or rows is
permissible.
You can use simulated data.
You only need to apply for IRB review if you are collecting data involving
human participants (this is very rare). Otherwise, your project is in IRB
compliance.
See a list of openly available data here.
Machine Learning requirements
You must apply machine learning to data.
application
Contents
What does the application need to do?
Coding
Application Performance
FAQ
Questions, comments, or suggestions?
Your software application (the “app”) is the entirety of part C of task 2.
However, the Computer Science capstone is not a software project. Other than
requiring you to apply machine learning (ML) to data, they will only assess
what your application does -not its presentation.
Tip
The official rubric and directions were written to map the project’s
elements to specific competencies. However, to allow for a broad
range of projects its language is necessarily also broad. For more
specifics, we recommend referring to the guidelines on this webpage
and the Task 2 template.
,What does the application need to
do?
Your app must help the user solve the organizational problem from task 1 by
doing the following:
1. Data → ML model: Use data to develop a machine learning model.
2. Accuracy Metric: Provide an appropriate metric or plan for measuring
the app’s accuracy.
3. Visualizations: Present 3 different pictures; images can code generated
or static.
4. User Application: Provide a way for the “user” to use the ML model
towards solving the problem.
Example 1 Example 2 Example 3 Example 4
The App: A standalone Jupyter Notebook file (.ipynb) which can predict Iris
types (see the example).
Data → ML model: Labeled Iris petal dimensions train an SVM
classification model.
Accuracy Metric: Percent of correct predictions on testing data.
Visualizations: Histograms showing distributions of different flower
features and a confusion matrix illustrating the accuracy of the
classification model.
User interface: (New data → ML model → prediction) Following
detailed instructions, the user can input flower dimensions, run code
in the Jupyter notebook, and the app returns a prediction helping
customers identify their flowers.
, Data requirements
Your data must be sufficiently complex and processed for your ML
algorithm to function as needed.
Evaluators must be able to access your data.
There are no specific requirements regarding the complexity or nature of your
data. However, your application must work and fulfill the organizational need
you’ll outline in the task 2 documentation. Therefore, when choosing your raw
data, consider carefully any additional time and effort necessary to prepare
that data for use. As these extra steps are only needed in so far as the chosen
problem and algorithm need them, you may wish to adjust your project or
choose a different data set accordingly.
You do not need special permission to use any open-source dataset (see our
list). Furthermore, data sets used in previous C964 projects are available for
reuse (no list of previously used datasets exists). It’s only required that the
data meet the needs of your project and be legally available to use and share
with evaluators. For example, using data belonging to a current employer
would require submitting a waiver form.
You can use data used for previous projects (submitted by you or others)
including any dataset found on kaggle.com.
You do not have to use all the data. Removing columns or rows is
permissible.
You can use simulated data.
You only need to apply for IRB review if you are collecting data involving
human participants (this is very rare). Otherwise, your project is in IRB
compliance.
See a list of openly available data here.
Machine Learning requirements
You must apply machine learning to data.
