Esmael Shaban

• Built and deployed Prophet, NeuralProphet, LSTM, and stacked models (SARIMAX + XGBoost) and XGBoost as META model , achieving MAE of 700—6x lower than baselines for 12-month forecasts (This accuracy was for stacked model) . • Combined linear and non-linear models with XGBoost as the meta-model, enhancing prediction accuracy. • Applied K-fold cross-validation to improve data utilization and model performance, driving revenue growth through accurate forecasts.

• Implemented K-means clustering for customer segmentation, increasing conversion rates by 22%. • Deployed fractal-based algorithms for complex pattern recognition, improving accuracy in spotting unique customer • behaviors. • Utilized T-SNE and PCA for dimensionality reduction, accelerating neural network training and achieving a 93% accuracy in customer segment classification.

- Built CNN-based classification (ResNet variants) for sun/cloud detection, achieving 96% accuracy - Enhanced temporal modeling with RNNs, improved decision-making process by 30%

Developed data analysis pipeline for stock market prediction, increasing project efficiency by 20%