Overview / Usage

A Recommender System predicts the likelihood that a user would prefer an item. Based on previous user interaction with the data source that the system takes the information from (besides the data from other users, or historical trends), the system is capable of recommending an item to a user. Think about the fact that Amazon recommends you books that they think you could like; Amazon might be making effective use of a Recommender System behind the curtains. This simple definition, allows us to think in a diverse set of applications where Recommender Systems might be useful. Applications such as documents, movies, music, romantic partners, or who to follow on Twitter, are pervasive and widely known in the world of Information Retrieval.

A collaborative filtering algorithm works by finding a set of people (assuming persons are the only client or user of a RS) with preferences or tastes similar to the target user. Using this smaller set of “similar” people, it constructs a ranked list of suggestions. There are several ways to measure the similarity of two people. It’s important to highlight that we’re not going to use attributes or descriptors of an item to recommend it, we’re just using the tastes or preferences over that item.

​

The future aspiration for this project is that, the course recommender system would be extended to the whole of CSS department. Once the recommender system is developed for the CSS department, it can quite easily be extended to the entire university and be integrated

Methodology / Approach

The data for the system comprised of old students records from the CSS department. The dataset contains 318 student and the grades they received for 30 different courses. This data is divided into 2 sets: training and testing. Training data set is used to build the Predictive model and the Test data is used to validate the prediction.  Fig: 1 shows an overview of the working of the predictive model. After a detailed study ,Collaborative filtering technique [Jannach; et al, 2011] using the K-nearest Neighbor (KNN) Machine Learning Algorithm was identified as the best suitable approach to build the predictive model. KNN is a method for classifying objects based on closest sample in the entire data set. In the case of Course Recommender system, KNN is used to identify and classify all the students who have chosen the same courses and received similar grades. Then, the Pearson Correlation coefficient (PCC) is calculated for each of the student pair, so as measure the similarity or dissimilarity between the 2 students. Depending on the PCC value, a weight is assigned to each sample. A weighted average of the all the student yield the predicted grade for a student, for a particular course.Both, the predictive model and the GUI for the Course Recommender system are developed using the Python programming language.This capstone project has been a tremendous learning experience. The following are some of the things that were learned over the course of the 2 quarters:

• Learned numerous Machine Learning algorithms
• Data acquisition and pruning
• Software development using Python Programming language and MATLAB.

Technologies Used

MatLab , R ,Python 2.7+, K-nearest Neighbor (KNN) , LSTM, RNN

Repository

https://github.com/ajinkyajawale14/Heratech_Hackathon/blob/master/model2.0.ipynb

Other links

• Ajinkya Jawale Fork (https://github.com/ajinkyajawale14/Heratech_Hackathon/blob/master/model2.0.ipynb)