Skip to content

nishnash54/RecOmax---Recommendation-Platform

Repository files navigation

RecOmax

Recommendation Platform

Instructions to RUN

  • Install Ipython (Jupyter Notebooks recommended)
  • Install requirements run pip install -r requirements.txt
  • For recommendation engine run Recommendation_engine.ipynb
  • For prediction engine run Prediction_engine.ipynb
  • Dashboard and notebooks Demo website

If you don't want to run the scripts, the same are available on the Demo website

Problem statement

  • Creating a product recommendation system
  • With the help of Machine Learning increase sales
  • Reduce time spent by Sales Representatives

Solution

  • Create a Product to Store based recommendation system that is Centrally controlled
  • Recommendations made based on previous sales history of the product
  • Integrate Online sales trends to provide better real time product recommendation

The vision

In today's world, data analysis coupled with the power of Machine learning and Artificial intelligence (deep learning) is helping companies solve the most complex of problems. We designed RecOmax as a ready to use platform that will help P&G predict sales of a specific item in a specific store based on historical sales data and complex trend analysis. We aim to build end to end solutions that benefit the client and provide them an edge over their competitors.

The build

The build can be divided into 3 main sections. These are the Recommendation engine, the Prediction engine and the Client facing data dashboard (report). These fields are elaborated on below

  • Prediction engine This engine works off a Kaggle data set. The main aim of the engine to is predict the sales of a specific item in a specific store location for the next month. The engine makes two types of predictions

    • When historical sales data is present, that is when we have sales data of a particular item from particular shop and
    • When historical sales data is absent, the engine analyses trends in item sales based on factors such as store locations, date-time features and various other measures. The prediction engine has a Root mean square error of 1.33.
  • Recommendation engine This engine works on the provided data set. It aims to recommend to the end user similar items based on the current selection. The data set provided looks minimal, but we used data analysis to generate features for each individual item from the given data. As this is an online data set, the trend analysis of this data is integrated into the Prediction engine also.

  • Data dashboard A fully interactive dashboard to present our final platform to the client with data visualizations and information regarding the working of various features in the platform. We used the professional Business Intelligence tool Power BI to prepare the client end report (dashboard)

Product roadmap

Overtime we have planned on making this project in to a full fledged tools with big data integration and real time online data integration. Using big data tool such as Hadoop and Mapreduce, we want to take this project one notch higher by computing tremendous amounts of data. Another plan is to based on the fact that online sales data is available almost immediately. This means that using that data will give the engine an edge over any older historical data prediction model.