This Repo is part of my diploma research where I tried to forecast time series with Deep Learning custom models

Time Series Forecasting with DL&Attention Builds LSTM+Attention and Transformer models to predict future values in multi-variate time series data. Implements custom attention layers for interpretability, performs hyperparameter tuning, and compares against baselines.

ML & DL-based Disaster Predictor that forecasts disaster severity using classification, regression, and time-series models. Includes Streamlit GUI, LSTM forecasting, and cloud-ready deployment. Built with Python, Scikit-learn, Keras, and more.

A hybrid time series project combining Deep Learning (LSTM) for forecasting with Machine Learning models for trend classification. Forecasts future values, extracts features, and classifies sequence patterns, integrating ML + DL for accurate and interpretable results.

Developed a stock price forecasting framework for Apple Inc. using both traditional and deep learning models. Implemented Linear Regression, LSTM, and a Stacking Ensemble with XGBoost, achieving improved accuracy and robustness in predictions. Demonstrated the effectiveness of combining ML and DL for financial time series forecasting.

Time-series energy consumption forecasting. Classifies building usage patterns via K-Means clustering, then predicts electricity demand using ML (CatBoost, LightGBM, XGBoost, RF, DT) and DL (1D-CNN LSTM, Seq2Seq) models with KMA weather and KIER usage data.

Multivariate air quality forecasting using time-series DL models to predict key pollutants from historical sensor + weather features, with end-to-end preprocessing, feature engineering, hyperparameter tuning, and evaluation (MAE/RMSE) for reliable short-term forecasts and long term forecasts.

Time-series forecasting pipeline comparing 5 ML/DL models (Ridge, ARIMA, Random Forest, XGBoost, LSTM) on UCI household power consumption data. XGBoost achieved R²=0.6055 with 17 engineered features including temporal patterns and lag-based history. Includes full preprocessing, EDA, feature importance analysis, and visualizations.

This project forecasts the flexible regulation capacity of EVs, showing how much energy they can return to the grid. Using an EV charging dataset, it compares time-series (ARIMA, Prophet), ML (XGBoost, LightGBM), and DL (MLP, LSTM, GRU) models, with a focus on probabilistic forecasting for smart grid planning.