recommendations-with-IBM

Table of Contents

1. Motivation
2. Requirements
3. Files
4. Different parts

Motivation

In this project which is a part of Udacity nanodegree program, four different recommendation systems to on real data from the IBM Watson Studio platform is studied:

• Rank Based Recommendation.
• User-User Based Collaborative Filtering.
• Content Based Recommendations.
• Matrix Factorization.

Packges

The project was implemented using Anaconda distribution of Python 3.0. Moreover I have used the following python libraries:

1. Pickle
2. Matplotlib
3. NLTK
4. NumPy
5. Pandas

File Descriptions

There is a jupyter note book file and a html file that implements the recommender engine.

Different parts

Exploratory Data Analysis(EDA)

Data cleaning, missing value imputation have been done in this part of this project.

Rank Based Recommendations

The most popular articles are found based on users interactions.

User-User Based Collaborative Filtering

Similarity between users are studied.

Content Based Recommendations

Different methods to find similarites between documents NLP techniques are used.

Matrix Factorization

Matrix decomposition is used to predict new articles an individual might interact with.