/pythontimeseries

Time series analysis using Python and benchmarking time series forecasts.

Primary LanguageJupyter NotebookMIT LicenseMIT

image info

SARChI - Python: Time Series Analysis

South Africa-Switzerland Bilateral Research Chair in Blockchain Technology aims to explore blockchain integrations with real-world applications and development in Agric-food.

Context

As a research center, our focus is more than just blockchain technology, but we have expertise in web development, data analysis, data science, and machine learning development. In this repository, look at a niche area of data science and machine learning engineering, time series forecasting. We use the fundamental gained from our Python fundamentals series, and apply it to Demand Forecasting.

Dataset

The dataset used for this demand forecasting problem is the dataset used in Global Energy Forecasting Competition 2014 (GEFCom2014).

Forecast Methods

For this problem, we forecast on performing time series analysis using Python and we introduce the following benchmark time series forecasts methods:

  1. Average Forecast
  2. Naive Method
  3. Seasonal Naive Method
  4. Drift Method

The goal is to build on this on later editions to create more advanced statistical forecasting methods and eventually compare them with the performance of machine learning solutions.

Resources