Key Lessons from Winning the 2025 China University Big Data Competition

Team Overview

Team Name: MMMM

Team Member: Man Li Chen (Jilin University)

Team Ranking: National First Place

Competition Experience and Technical Decisions

During the competition, I realized the importance of balancing model performance with runtime efficiency under multiple real‑world constraints. To avoid the unpredictable nature of the stock market, I simplified the problem by focusing solely on the competition’s target task and eliminating high‑uncertainty components.

Instead of attempting to predict the exact future price change of each stock—a task that is extremely complex and influenced by many uncontrollable factors—I adopted a ranking‑learning approach. This method does not aim for absolute values but learns the relative strength order of stocks, allowing flexible adjustment of bin numbers to suit different granularities.

To handle the asymmetry between rising and falling stocks, I designed an ensemble model consisting of two independent modules, each dedicated to identifying the top‑10 stocks with the greatest upward or downward movement.

Feature Engineering

Feature engineering proved crucial. I incorporated a rich set of technical indicators and candlestick patterns, and constructed both individual‑stock features and sector‑level linkage features that capture co‑movement effects. These features significantly enhanced the model’s ability to perceive collective market behavior, improving performance from both individual and systemic perspectives.

Through ablation experiments and feature‑importance analysis, I confirmed the effectiveness of the modeling strategy and the engineered features.

Reflections

The competition reinforced the need to carefully balance advanced technical solutions with practical implementation constraints. True innovation must be grounded in real‑world scenarios, requiring a systematic mindset that weighs objectives, resource limits, and evaluation criteria to find an optimal equilibrium.

I am grateful to the organizers for providing a valuable platform that fostered high‑level technical practice and personal growth. The experience deepened my understanding of the essence of technology and its application to real problems, which I will carry forward into future research and engineering work.