/HackerEarth_Hackathon_Slashing_Prices_Prediction_for_Sale

Secured `5th` position in this Hackathon

Primary LanguageJupyter Notebook

HackerEarth_Hackathon_Slashing_Prices_Prediction_for_Sale

Secured 5th Rank in this Competition

Problem Statement

A leading global leader of e-commerce has over 150 million paid subscription users. One of the many perks of the subscription is the privilege of buying products at lower prices. For an upcoming sale, the organization has decided to promote local artisans and their products, to help them through these tough times. However, slashed prices may impact local artists.

To not let discounts affect local artists, the company has decided to determine the lowest price at which a particular good can be sold. The task is to build a predictive model using Machine Learning that helps them set up a lowest-pricing model for these products.

Approach

  • Removed skewness from ['Demand', 'High_Cap_Price', 'Low_Cap_Price'] columns using log transformation

  • One-Hot representation of categorical columns

  • Normalizing certain numerical features

  • Final Submission : Ensembling various regression models {

                              `XGBoost`,
                          
                          `Light Gradient Boosting Model`,
                          
                          `Gradient Boosting Model`
                          
                          }

  • [Didn't work] Pseudo Labeling on Test Set

  • [Didn't work] Label Encoding of categorical features