This is a on-going Kaggle project to predict grocery sales using machine learning. Various time-series analysis and exploratory data analysis are performed to catch the key indicators. A hybrid model is ensembled for sale prediction.
After stepping into the data science world, I heard one world many times, time-series forecasting. It is a very power technique to provide useful information to stake-holders in order to make strategic decisions. Therefore, I took the time-series study course on Kaggle and put the knowledge I learned into practice in this project.
An accurate model to predict sales for up-coming months is crucial to a grocery store as you can image. It can help reduce issues related to overstocking and improve customer satisfaction, which are key to the success of any retail store. With sales data from Corporación Favorita, a large Ecuadorian-based grocery retailer, during the time period from Jan 2013 to Aug 2017, how much were sold in the rest months in 2017 are the questions to answer by competitors in this challenge.
- Create a model to accurately predict sales of up-coming months
- Answer the question: What factors impacting the grocery sales most?
- Python for:
- time-series data analysis
- EDA of sales data
- seasonal pattern exploration
- indicator generations
- machine learning model building
- prediction evaluation
- A machine learning hybrid model combines:
- Linear Regression
- XGBRegressor
The data is published on Kaggle.