JForecast — Making Forecasting Simpler & Better

JForecast is an advanced, cutting-edge and robust forecasting tool that leverages Java and Weka library's power. It offers a diverse range of functionalities tailored to streamline forecasting processes, ensuring accurate predictions for both simple time series and correlation forecasting. JForecast SDK significantly simplifies the entire process while optimizing memory usage & performance. JForecast distinguishes itself as the exclusive Java SDK which provides state-of-the-art (SOTA) support for correlational forecasting.

Features

• Built-in Evaluators: Automatic dataset splitting based on specified percentage, supporting evaluation metrics like MAPE and RMSE
• Flexible Prediction Options:
  • Prediction by variable name.
  • Prediction by units ahead.
  • Prediction by end date, based on defined periodicity.
• Simple Syntax: Intuitive and concise syntax for seamless usability.
• Data Format Support:
  • Supports both ARFF and CSV formats.
  • Automatic sorting of data in ascending order based on the defined dateTime column.
• Integration of XGBoost: Incorporate state-of-the-art algorithms for superior forecasting accuracy.
• Correlation Forecasting:
  • Automatic Data Preprocessing ‐ Prepares datasets based on specified predictor and target variables.
  • Imputation ‐ Handles missing values, addressing the inconsistency challenge in observations with a robust approach
  • Data Transformation ‐ Offers four data transformation approaches: Yeo-Johnson Transformation, Box-Cox Transformation, Log Transformation, and Standardization. These approaches enhance interpretability by reducing skewed distributions among variables, crucial for improving data normality and model accuracy.
  • Feature Engineering & Prediction ‐ Captures complex relationships, condenses information, and performs correlated predictions using specified algorithms.