CrowdAnalytix: Predicting likelihood of Online purchases
The goal of this modeling contest is to predict the likelihood of online purchases using visits and purchases ecommerce data. Prizes for the Top 5 contestants on the private test set leaderboard.
A peek in to online shopper's behaviour is very insightful. What do they do online? Where does all the time get spent? If we attempt to answer these questions, we would see that most of consumers time is spent researching about products or reading expert and user reviews. Rest of the time is spent in price comparison sites and searching for coupons. These facts suggest a complex behaviour of consumers purchase decision. The diversity of products available in the online marketplace makes it even more complex. Hence the main objective of this contest is to predict the likelihood of online purchases by consumers.
The goal of this contest is develop a model to predict customer likelihood to make a purchase or not, based on the given features. Feature description explained in section below. The true label of whether a visit becomes a purchase or not is provided as "Outcome Label" in the training dataset.
Addtional objective of the contest is to present a PDF report analysing features and insights based on modeling performed. The report has to be precise, should be based on facts from modeling excercise and should try to answer some of the following questions, but not limited to:
- Features more representative of purchase or not.
- How can additional features affect the model performance.
- Modeling Approach & Why?
- Running time complexity of the model
Learn more and participate at