README.md

NLP-Classifiers-Analysis

Description

We are using A bag of words algorithm to predict the sentiment analysis on a resturant's reviews dataset. Now after getting the processed data from a bag of words algorithm, we are applying different types of classifiers on the data and analyse the performance of the classifiers on this dataset.

Dependencies and Libraries Used:

1. Pandas
2. Scikit-learn
3. nltk library

Please download stopwords before executing the related operation Download it by running the command: >>nltk.download('stopwords')

Classifier Models Used:

1. Logistic Regression

2. Naive Bayes

3. k-NearestNeighbors

4. Support Vector Machine (SVM)

5. Decision Tree Classification

6. Random Forests Classification

7. Gradient Boosted Decision Tree Classification

Improvement Methods Used:

kCrossValidation - for more accurate and random Testing of model

GridSearch - for tuning the Hyper Parameters

Result and Analysis:

Naive Bayes:

Accuracy = 67.37 %

Standard Deviation = 0.0499

Precision = 0.567

Recall = 0.821

F1 Score = 0.6707

k-NearestsNeighbors:

Best Config: n = 5, p = 2 (Euclidean)

Accuracy = 65.87 %

Standard Deviation = 0.02354

Precision = 0.6316

Recall = 0.6186

F1 Score = 0.625

Decision Tree Classification:

Accuracy = 75.35 %

Standard Deviation = 0.0433546

Precision = 0.7368

Recall = 0.6796

F1 Score = 0.70705

Random Forests Classification:

Best Config: n = 73, criterion = 'entropy'

Accuracy = 77.24 %

Standard Deviation = 0.0201

Precision = 0.7684

Recall = 0.6822

F1 Score = 0.72274

Gradient Boosted Decision Tree Classification:

Best Config: loss = 'deviance', n = 113

Accuracy = 77.5 %

Standard Deviation = 0.0251

Precision = 0.8211

Recall = 0.67241

F1 Score = 0.7394

Logistic Regression

Accuracy = 78.13 %

Standard Deviation = 0.0205

Precision = 0.7368

Recall = 0.7071

F1 Score = 0.7217

Support Vector Machine (linear)

Accuracy = 78.38 %

Standard Deviation = 0.0202

Precision = 0.7579

Recall = 0.6923

F1 Score = 0.7236

Recommended Classifiers for the Model:

1. Support Vector Machine

2. Logistic Regression

3. Gradient Boosted Decision Tree Classification

4. Random Forests Classification

Contribute to this repository:

Fork this repo and download the dataset (Restaurant_Reviews.tsv) and the template (Contribution Template) and insert your classifier after the Comment ("#INSERT HERE"), use the GridSearch accordingly for your classifier and rename the template to the name of the classifier and upload it and then create a pull request.