Time Series

  • Pre-built Models: Includes popular time series forecasting models like ARIMA, SARIMA, torch, and more.
  • Seamless Integration: Load data, preprocess, and run forecasting models in one place.
  • Automatic Forecasting: Automatically generates forecasts for future time steps once a model is selected.
  • Visualization: Built-in tools for visualizing both historical data and forecasted values.
  • Customizable: Fine-tune model parameters to suit your specific use case.
  • Extensive Documentation: Detailed guides and examples to help you get started.

Import SimpliML Time Series

import pandas as pd
import simpliml.timeseriesforecast as tsf

Build Time Series Data

sourceDF = pd.read_csv("") # Any Input data read
dataDF, futureDF = tsf.generateTSData(sourceDF, format='%Y-%m', freq='MS', periods=30)

Parameters:-

Analysis Data

tsf.analysisData(dataDF) # This will work only in interactive computational environment like Jupyter Notebook/lab/hub ..etc

Build Model and forcast

mdlResult = tsf.runModel(dataDF, futureDF, seasonal=12, modelApproach = 'FAST', testSize=80)  # Single Process Thread
(OR)
mdlResult = tsf.runThreadModel(dataDF, futureDF, seasonal=12, modelApproach = 'FAST', testSize=80) # Single Process Multiple Thread (Thread : CPU Count * 2)
(OR)
mdlResult = tsf.runProcessModel(dataDF, futureDF, seasonal=12, modelApproach = 'FAST', testSize=80) # Multiple Process (Process : CPU Count / 4) # Advise to use only in Windows

Parameters:-

  • seasonal : int, optional
    • The number of periods in a complete seasonal cycle,
    • Example
    • 1 : Yearly data
    • 2 : Half-yearly data
    • 4 : Quarterly data
    • 7 : Daily data with a weekly cycle
    • 52 : Weekly Data
  • modelApproach : ["BEST", "FAST"], optional
    • BEST : Best model build with multiple permutation and combination
    • FAST : Fast model build with limited permutation and combination
  • testSize : int, optional
    • Test Size by defult 80:20 rule

Model Result Analysis

mdlOutPut = tsf.modelResult(dataDF, mdlResult, modelApproach='Best')

Parameters:-

  • modelApproach : str, optional
    • BEST MAPE analysis report and can pass the model name, get the analysis report