adriahabib/aunet_timeseries

AUNET: Attention-Based Time Series Forecasting

AUNET is a neural network architecture designed for time series forecasting, combining multi-head self-attention with dense layers to capture temporal patterns in numeric datasets. Developed using TensorFlow/Keras, it supports customizable input windows and forecast horizons.

Usage

from aunet_model import AUNET

model = AUNET(input_length=30, forecast_horizon=7)
model.fit(X_train, y_train)
y_pred = model.predict(X_test)

Model Details

• Architecture: Multi-head self-attention + dense layers
• Framework: TensorFlow / Keras
• Use Case: Multistep forecasting of univariate or multivariate numeric time series
• Target Feature: Second column of data (post date-drop)

Authors

• Adria Binte Habib
• Dr. Golam Rabiul Alam
• Dr. Zia Uddin