TensoCast - Traffic Forecasting
Tensor decomposition + LSTM/PyTorch + XGBoost baselines in one comparative pipeline.
A research / forecasting project exploring how tensor decomposition can capture spatial-temporal structure before downstream predictive modeling. The work combines Python experiments, data preprocessing, model comparison, and forecasting evaluation.
Combined decomposition and predictive models into one pipeline
Improved results over classical baselines
Strengthened my research-to-implementation workflow
Impact
35% improvement over baseline approaches
Role
ML / Forecasting Project
Timeline
2024
Key tags
Problem
Traffic data carries rich spatial and temporal structure that many baseline approaches fail to model effectively.
Solution
Used tensor decomposition to represent multidimensional traffic dynamics, then layered predictive models on top for forecasting.
Architecture
Tensor preprocessing and decomposition feed learning models built with PyTorch and classical ML components, with experiments comparing performance across modeling choices.
Challenges
- Choosing decomposition techniques that preserve useful structure
- Balancing interpretability with predictive performance
- Handling noisy and incomplete sensor data
Technology stack
Continue exploring
More case studies
Lucid - AI Notes Platform
Helps students turn scattered lecture screenshots and notes into searchable study material.
Knowledge Graph RAG (ToPWR)
Helps students and staff query university information spread across pages, departments, and systems.
NeuroPark - 3D Vehicle Detection
Supports smart parking analysis by estimating vehicles and scene geometry from regular RGB images.