APIForecast

The purpose of this project is to demonstrate the application of three main forecasting functions: single exponential smoothing, double exponential smoothing and Holt-Winters forecasting.

Prerequisites

Matplotlib

Install Matplotlib with pip install matplotlib

NumPy

Install NumPy with pip install numpy

PyShark

Install PyShark with pip install pyshark

SciPy

Install SciPy with pip install scipy

Description

The project consists of a few Python files:

• APIForecast.py: here are implemented the three forecasting functions, along with SSE and RSI functions and two fitting functions. The fitting functions differs in the algorithm that they use: one uses the Nelder-Mead algorithm with some tweaks, and the other uses the TNC fitting algorithm from the SciPy package.
• Utils.py: this files contains the plotting functions, which uses matplotlib, and a couple of utility functions for output formatting.
• CreateDatasets.py: it implements the generation of the values datasets which we used for testing, with the options to create a "normal" dataset or an "anomalous" dataset. It also can produce a dataset from a pcap file.
• Three demo scripts:
  • Demo.py: used to test the API by passing what we want to do via arguments on the CLI. It can read both json datasets and pcap files.
  • DemoInteractive.py: and interactive version of the previous script.
  • Test.py: an automatic test for Holt-Winters with default parameters.

Usage

First, create a dataset with python3 CreateDatasets.py --type series --days 5

Then, we can use Demo.py to do a forecasting demo: python3 Demo.py --dataset dataset.json --season 288 --rsi 24 --alpha 0.57300 --beta 0.00667 --gamma 0.92767

Or we can use Test.py to do another forecasting demo: python3 Test.py