nhull/logistic-regression-model

Logistic Regression Sentiment Analysis Model

This model is a Logistic Regression classifier trained on the TripAdvisor sentiment analysis dataset. It predicts the sentiment of hotel reviews on a 1-5 star scale. The model takes text input (hotel reviews) and outputs a sentiment rating from 1 to 5 stars.

Model Details

• Model Type: Logistic Regression
• Task: Sentiment Analysis
• Input: A hotel review (text)
• Output: Sentiment rating (1-5 stars)
• Trained Dataset: nhull/tripadvisor-split-dataset-v2

Intended Use

This model is designed to classify hotel reviews based on their sentiment. It assigns a star rating between 1 and 5 to a review, indicating the sentiment expressed in the review.

---

The model will return a sentiment rating between 1 and 5 stars, where:

• 1: Very bad
• 2: Bad
• 3: Neutral
• 4: Good
• 5: Very good

---

Dataset

The dataset used for training, validation, and testing is nhull/tripadvisor-split-dataset-v2. It consists of:

• Training Set: 30,400 reviews
• Validation Set: 1,600 reviews
• Test Set: 8,000 reviews

All splits are balanced across five sentiment labels.

---

Test Performance

Model predicts too high on average by 0.44.

• Test Accuracy: 61.05% on the test set.

• Classification Report:

Label	Precision	Recall	F1-score	Support
1.0	0.70	0.73	0.71	1600
2.0	0.52	0.50	0.51	1600
3.0	0.57	0.54	0.55	1600
4.0	0.55	0.54	0.55	1600
5.0	0.71	0.74	0.72	1600
Accuracy	-	-	0.61	8000
Macro avg	0.61	0.61	0.61	8000
Weighted avg	0.61	0.61	0.61	8000

• Confusion Matrix:

True \ Predicted	1	2	3	4	5
1	1165	384	41	3	7
2	432	805	315	31	17
3	61	314	857	311	57
4	3	48	264	870	415
5	6	10	32	365	1187

---

Files Included

• validation_results_log_regression.csv: Contains correctly classified reviews with their real and predicted labels.

---

Limitations

• The model performs well on extreme ratings (1 and 5 stars) but struggles with intermediate ratings (2, 3, and 4 stars).
• The model was trained on the TripAdvisor dataset and may not generalize well to reviews from other sources or domains.
• The model does not handle aspects like sarcasm or humor well, and shorter reviews may lead to less accurate predictions.