Skip to content

Files

Latest commit

 

History

History

Daimler Forecasting

Forecasting

This repo contains codes and report created for time series competition by Daimler.

Problem Statement

Analysis and forecasting of two different time series with daily frequency.

Techniques

  1. Visualization and analysis of time series - data_visualization.py
    This includes yearly plots, monthly plots, lag scatter plots, autocorrelation plots and stationrity checks.
  2. Baseline forecast - baseline.py
    MSE error was calculated by forecasting with lag 1.
  3. SARIMA - seasonal.py
    Implementation of Seasonal ARIMA in python.
  4. AR - autoregression.py
    Implementation of Autoregression (AR) models in python.
    Link for reference - https://machinelearningmastery.com/autoregression-models-time-series-forecasting-python/
  5. Prophet - prophet.py
    Implementation of prophet in python. This performed best in this case for both time series.
    Transformation and parameters are different for both time series.

Report

Final report - report.pdf
This report includes, detailed approach and final forecasts with reasonings.