For detailed insights and visualization of the data, my inferences, modelling, Prediction please visit my BLOG.
- Learning and Applying data wrangling, data exploration and forecasting methods.
- Applying various forecasting methods and coming up with better predictions for 28-100 days for all the items in the data.
- Learn to tell a story and practice various data visualization techniques
- WALMART data from three states California, Texas, Wisconsin for three product Foods, household and hobbies.
-
For EDA - see COLAB
-
LightGBM - This notebook contains the implementation for Boosting technique LightGBM to forecast time-series models.
- 9 different time-series were explore instead of 42000 time series
- Different forecasting methods have been experimented and explored.
- RMSE metric returned about 1.5 - 2.5 for 9 time series models.
- Explore LSTM, Prophet
- Explore more time series, finding trend and seasonality between items in the products and combine them to explore more time-series