yao-jz / reviews-rating-prediction

A simple Python implementation with ensemble learning to predict rating according to reviews.

Reviews-Rating-Prediction

Using an ensemble learning approach, bagging and adaboost algorithms are implemented to perform the classification.

Implementation Details

Feature Extraction

Using the tfidf vectorization method to do the feature extraction for the content part of each data to obtain a vector for each comment text representation, and the vector is saved as an npy file for svm and decision tree training.

Model

SVM and decision tree, using LinearSVC and decision tree models provided by sklearn for training and prediction respectively.

Ensemble Learning

I implemented the Bagging and Ada Boost algorithm.

Experiments

Model	MAE	RMSE	MSE
Baseline SVM	0.6015	1.074011173126239	1.1535
Baseline DTree	0.9195	1.4281106399715675	2.0395
Bagging SVM	0.56725	1.0560539758932779	1.11525
Bagging DTree	0.72275	1.308147545195113	1.71125
Ada Boost SVM	0.69325	1.2687592364195817	1.60975
Ada Boost DTree	0.72275	1.3085297092538632	1.71225