Predicting Future Retail Sales for 1C Company

Kaggle | Machine Learning Competition
13th place (RMSE: 0.94X) in December, 2017



Project Description

The 1C Company : Predict Future Sales competition is a reoccurring playground prediction challenge on Kaggle.

Skills

Data Analysis, Data Cleaning, Regression, Time-Series / Panel Data, Feature Interactions & Selection, Ensemble Methods, Generalized Stacking

Tools

Python, Pandas, NumPy, Matplotlib, Jupyter, Seaborn, XGBoost, LightGBM, Random Forest / Extra Trees, Linear Regression, ElasticNet, Artificial Neural Network (Dense), AdaBoost


Motivation

The 1C competition is meant to serve as a playground for dealing with various concepts of time-series regression, handling the interpolation of missing data, and discriminitive filtering of unnecessary information. Top-ranked submissions will have to deal with seasonality, ARIMA, selecting P and Q using AIC, and other common techniques in working with prediction from panel data.



See the complete project in my GitHub repository.