SooyeonWon / time_series_analytics

Time Series Forecasting Models: ETS, ARIMA

Time Series Forecasting of Sales

by Sooyeon Won

Keywords

• Analytical Framework
• Time Series Visualisation
• Comparison the results of ETS Models vs. Seasonal-ARIMA
• Model Validation
• Time Series Forecasting

Introduction

This analysis is the project for "Time Series Forecasting" in Udacity Predictive Analytics Nanodegree Program. The goal of the project is to forecast monthly sales data for a video game company, in order to help plan out the supply with demand for the company's video games Initially, I conducted the analysis using the recommended software; Alteryx. However, for this analysis, I mainly went through the same project using python. Though the values of the findings are not exactly matched with the outputs based on Alteryx, I could reached very similar results.

Summary of Findings

This analysis is mainly about forecasting for upcoming sales in a video game company. Firstly, I investigate and prepare the time series data. The provided data was appropriate to use time series models and I held out the last 4 periods of data points for validation. Then, I determined Trend, Seasonal and Error components in the data based on decomposition plots. After that, I analyse the data by applying the ARIMA and ETS models and describe the errors for both models. I compared the in-sample error measurements to both models and compare error measurements for the holdout sample in the forecast. Finally,I choose the best fitting model and forecast the next four periods.

References

• MASE
  
• RMSE, MAE
  
• MASE
  
• ETS
  
• Seasonal ARIMA reference
  
• Decomposition