WGU D604 Task 2: Sentiment Analysis with Neural Networks | Complete Project Report, Bi-LSTM Model & Python Code | 2026/2027 Updated

WESTERN GOVERNORS UNIVERSITY | D604



Advanced Analytics
Task 2
Sentiment Analysis with
Neural Networks



Project scenario: binary customer-review sentiment classification using the UCI Sentiment Labelled
Sentences dataset and a bidirectional long short-term memory (Bi-LSTM) network.

Technical coverage: research question, dataset exploration, preprocessing, tokenizer and padding,
network design, backpropagation, stopping criteria, metrics, ethics, recommendations, executable code,
and sources.

Prepared: October 27, 2026 | Course: WGU D604 Advanced Analytics




Evidence and verification. This report is grounded in the eight-page Task 2 document provided by the user and official
public dataset and framework documentation. The provided report contains conflicting split sizes, parameter counts and
performance results. No TensorFlow training run was performed while producing this PDF. Historical reference-run figures
are identified explicitly. The full accompanying Python program can generate authentic evidence when executed with the
UCI dataset; a passing grade is not guaranteed.




Sentiment analysis with bidirectional LSTM | technical report 1

, WGU D604 / ADVANCED ANALYTICS / TASK 2




Contents and requirements coverage

Contents and requirements coverage 2


Executive summary 3


Part I. Research question and business problem 3

A1. Research question . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

A2. Objectives and success criteria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .4

A3. Neural network selection and rationale . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .4


Part II. Data preparation 4

B1. Exploratory data analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .4

B2. Text cleaning and tokenization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

B3. Sequence padding and truncation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

B4. Sentiment categories and output encoding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .6

B5. Split method and leakage controls . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

B6. Prepared dataset artifacts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7


Part III. Neural network architecture 7

C1. Network summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

C2. Layer depth, types and number of units . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

C3. Loss function, optimizer, activations and stopping criteria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8


Part IV. Neural network model evaluation 9

D1. Early-stopping behavior and reported training outcomes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

D2. Model fitness, variance and bias . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

D3. Curves and diagnostic evidence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

D4. Test performance, confusion matrix and consistency checks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

D5. Ethical AI, fairness, privacy and business risk . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10


Part V. Summary and recommendations 11

E. Saving and reloading the neural network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

F. Network functionality and constraints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

G. Recommended course of action . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11


Part VI. Reporting and reproducibility 12

H. Code, documentation and generated deliverables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

I. Source attribution for code and libraries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

J. References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

K. Professional communication and material limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13


Appendix A. Evidence reconciliation register 13




Sentiment analysis with bidirectional LSTM | technical report 2