← Back to About

About / Project

Project

Selected data and machine-learning projects.

2026 · Tianchi Competition

Fund Flow Forecasting

Forecasting baseline

Built an end-to-end forecasting baseline from five raw CSV files to 30-day purchase and redemption forecasts.

Organized four business tables in SQLite and assembled data with SQL joins.
Created calendar, holiday-position, and weekday-pattern features.
Used rolling monthly validation and forward feature selection.

2026 · Machine Learning

Air Quality Classification

Clustering Analysis Lead

Compared K-Means, hierarchical clustering, and Gaussian mixture models on 9,358 hourly sensor records.

Evaluated clustering with Silhouette, Davies–Bouldin, ARI, and NMI.
Used a cluster-disjoint split to reduce leakage between similar samples.
The team achieved a Macro-F1 score of 0.8328.