hangfei/machienlearning

MachienLearning

##Description Predicted the rating of Yelp reviews given review text and other review information

• Processed the raw data, including information retrieval, feature selection, data rescaling, bigrams
• Implemented and combined various models, including PCA/SVD, SVM, Logistic Regression
• Visualized sentiment analysis results
• Enabled restaurants to improve their services via the feedback of low rating review

##Project Architectrue (For detailed instructions, please check README, not README.md)
We have tried multiple methods to pushing the edges of our results. The optimal one(with best accuracy and best speed) is in the init_model.m and make_final_prediction.m file(You could try to run script_test.m to test).
For other models we have develop, please check the README file.
Data
The data is in the data package, including raw data and some other processed data.

####Results: RMSE: 0.89 (Off by at most half a star of the actual result for a review.)

####Supporting Package:

• Libnear(http://www.csie.ntu.edu.tw/~cjlin/liblinear/)

##Installation instructions Matlab, and related packages

##Where to get help TBD

##Contribution guidelines TBD

##Contributor list Hangfei Lin, Di Mu, Sanjay Paul

##Credits, Inspiration, Alternatives

---

####Infographics for key words of high rating and low rating. Credit: Di Mu