Business Improvement Recommender System
Goals:
- In this project we leverage the YELP dataset provide a faster approach that combines different machine learning techniques to extract most frequently discussed negative topics/keywords that will provide business with insights regarding what they should be doing right based on what they are currently doing wrong.
- We use sentiment analysis to extract features from negative reviews to get higher business specific accuracy.
Implementation:
- Using Neural Networks (LSTM, CNN) and SVM, a Sentiment Analysis module was developed to separate positve and negative yelp reviews for a specific business
- Topic modelling done on the negative reviews to extract latent topics using Latent Dirichlet Allocation
- Extracted the most common keywords among the topics and re-iterated the topic modelling for those keyword specific texts in the negative reviews to filter out false positives.
- Top 5 Final recommendations were given to business ordered by the most negative aspect that needs improvement
- The program was scaled up to handle multiple outlets for a specific business eg: Starbucks
- All backend data handling tasks leverage MongoDB for storing intermediate and data.
Output:
- A list of features/keywords that reflect the business aspects which need improvement.
Latent Semantics Based Movie Recommender System
Goals:
A movie recommender system that takes user feedback to improve the recommendation. The system is developed using SVD matrix factorization and optimized using ALS method
Implementation:
-
A program which, given all the information available about movies a given user has watched, recommends the user 5 more movies to watch, using SVD.
-
The order in which the movies are watched and the recency is taken into account.
-
The result interface allows the user to provide positive and/or negative feedback for the ranked results returned by the system to enable Task Relevance feedback.
-
Relevance feedback loop is implemented to improve the accuracy of recommendation.
Output:
- The system outputs the revisions it suggests.
- User feedback is taken into account (either by revising the query or by re-ordering the results as appropriate) and a new set of ranked results are returned.
Machine Learning Repository for Different Datasets
Goals:
To try different Machine Learning algorithms on datasets to get better understanding of their implementation and insights in performance
Implementation:
Each Dataset has its own directory which states the different algorithms used and their comparisons.
Output:
- Error plots
- Behavior for different parameter values