Worked Projects

Tripper - the tour guide of the world 🌍

• User-friendly platform that allows tourists to easily match with local guides🧳/ reshape the way to make new frinds
• AI-powered interaction to provide real-time, knowledge-rich interactions and plan. 🤖
• A social forum where users can share and discuss their travel experiences, fostering community interaction and information sharing. 🗣️

Bitcoin Price Prediction

This study aims to create an accurate sequence prediction model to forecast Bitcoin price trends. The publicly available dataset is from Kaggle, covering Bitcoin price data from May 5, 2013, to October 23, 2022.

Data Processing and Feature Engineering

Data Processing: Cleaned data, handled missing values and anomalies, converted timestamps to datetime format.

Feature Engineering: Calculated VWAP, selected monthly data to reduce noise, performed stationarity tests, and STL decomposition.

Close price against time

Dayly, Monthly, Seasonally and Yearly

Seasonal and Trend decomposition visualization

Model Development

ARIMA Model

Implemented time series prediction through stationarity tests and parameter optimization

Parameters Comparison

MSE, RMSE, MAE statistical histogram of 7 models

LightGBM: Improved training efficiency and prediction accuracy using histogram optimization and leaf-wise growth strategy.

XGBoost: Achieved high accuracy predictions by combining gradient boosting and regularization.

Transformer: Processed complex sequences efficiently through self-attention mechanisms and parallel processing.

Prophet: Modeled seasonality and trends in time series data.

LSTM: Captured long-term dependencies in time series data using long short-term memory networks.

SVM: Implemented classification and regression predictions through kernel functions and support vectors.

Model Comparison

Deep Learning Models: Performed best, effectively capturing long-term dependencies.

Seasonal Models: Handled seasonality and trends well.

Traditional Machine Learning Models: Struggled with long-term dependencies.

🔧 Tech Stack

Data Processing: Python, Pandas

Feature Engineering: VWAP, STL Decomposition

Machine Learning Models: ARIMA, LightGBM, XGBoost, SVM

Deep Learning Models: Transformer, LSTM

Time Series Models: Prophet

Libraries: scikit-learn, statsmodels, lightgbm, xgboost, pytorch, fbprophet

Evaluation Metrics: MSE, RMSE, MAE

Comprehensive Detailed Design& Research for a 22-story Podium Single Tower 🏢 Structure

• A complete Design Package including Design& Research Report, Engineering Drawings, and Calculation Report.
• Shear Core Design Assessment ensuring deflection and drifts at each level met Serviceability Limit State requirements.
• Structural Model Validation through comparison of 2D hand calculation results with 3D Finite Element Analysis outcomes.

🏢Sustainable Design and Analysis of a Nine-Story Steel Frame Office Building

• Innovative Structural Solutions: Design of composite slabs, beams, and a cantilever truss system to address site-specific challenges.
• Robust Steel Framework: Detailed analysis and design of steel columns and wind bracing for optimal stability and resilience.
• Precise Connection Design: Meticulous engineering of beam-beam, beam-column, and bracing connections for structural integrity.
• Comprehensive Analysis: Parametric study to explore alternative approaches and drive future enhancements.
• Integrated Documentation: Extensive reporting with detailed drawings and calculations, ensuring transparency and precision in every aspect of the design.

Analysis of Hexagonal and Triangular Lattice Structures in MATLAB

• Simulated different loading scenarios for each model in MATLAB and Excel.
• Validated model accuracy with proposed refinements.

Still a lot.......I will list when I got time