📦 Time Series Library in Java (tslib)

Overview

This repository provides a modular Java library for time series analysis and forecasting. It includes implementations of key smoothing algorithms, moving averages, data transformation tools, and stationarity testing — all essential for modeling and analyzing sequential data in domains like finance, economics, and engineering.

---

🚀 Features

• 📈 Core time series models: Exponential Smoothing, Moving Averages
• 🔁 Support for trend and seasonality: Single, Double, and Triple Exponential Smoothing
• 🧪 Stationarity testing with Augmented Dickey-Fuller (ADF)
• 🔄 Built-in utilities for transformation (log, sqrt, Box-Cox)
• 📊 Statistical summary utilities: autocovariance, ACF, PACF, mean, std, and more
• 🧹 Clean architecture with extensible interfaces

---

🧠 Core Components

📊 Statistical Utilities

• tslib.stats.Stats: Utility methods for calculating mean, variance, autocovariance, ACF, PACF, and more.

---

🔄 Data Transformation

• tslib.transform.Transform: Preprocessing methods for log, square root, cube root, and Box-Cox transformations.

---

📈 Moving Average Models

• tslib.movingaverage.SimpleMovingAverage: Fixed-window SMA for smoothing time series.
• tslib.movingaverage.CumulativeMovingAverage: Real-time CMA update of the running mean.
• tslib.movingaverage.ExponentialMovingAverage: EMA implementation with decay factor.

Each model implements the shared MovingAverage interface for consistency.

---

🔁 Exponential Smoothing Models

• tslib.model.SingleExpSmoothing: Single Exponential Smoothing (level only).
• tslib.model.DoubleExpSmoothing: Double Exponential Smoothing (Holt's method – level + trend).
• tslib.model.TripleExpSmoothing: Triple Exponential Smoothing (Holt-Winters – level, trend, seasonality).

All models implement the ExponentialSmoothing interface.

---

🧪 Stationarity Testing

• tslib.tests.AugmentedDickeyFuller: Java implementation of the ADF test, used to check for unit roots in time series data.

---

📂 Example Usage

List<Double> data = Util.ReadFile("data/hotel.txt");

ExponentialSmoothing model = new TripleExpSmoothing(0.5, 0.3, 0.2, 12, false);
List<Double> forecast = model.forecast(data, 5);

MovingAverage sma = new SimpleMovingAverage(3);
List<Double> smoothed = sma.compute(data);

double lambda = BoxCox.lambdaSearch(data);
List<Double> transformed = BoxCox.transform(data, lambda);