Entries from 2026-01-01 to 1 year
Summary -Content: Tried to improve performance through data understanding-Conclusion: Got stuck on Target Encoding Summary Goal Current Status and This Session’s Challenge Prompts and Chat Log Organizing Data Overview Causal Relations…
Summary -Content: Verified the differences from a high-performance public notebook-Conclusion: Performance improved by using sample_weight Summary Goal Current Status and This Session’s Challenge Prompt and Chat Log Extracting Differe…
Summary - Content: Built an ensemble model for a new competition- Conclusion: Completed parameter tuning for individual models and weight tuning for the ensemble Summary Goal Current Position and Challenges Prompts and Chat Log Overal…
Summary of This Article -Content: Tried clustering as one of the features-Conclusion: Clustering and other preprocessing steps do have an effect, but model selection has a much larger impact on performance Summary of This Article Goal f…
Summary - Content: Created and tested new features- Conclusion: Performance did not improve, but I feel like I understood the concept① Creating features by combining variables in advance can help the model learn② Simple combinations resul…
Kaggle Challenge Log #6 – Applying Data Understanding Playground S6E3 Day 2 “Predict Customer Churn”
Summary of This Article -Content: I tried splitting the model based on the values of highly influential variables.(Example: When MonthlyCharges differs between monthly payment and yearly payment, split the data accordingly and train separ…
Summary - Content: I asked Copilot what to do for data understanding in a Kaggle competition - Conclusion: The path for data insights and model integration has become clear① Identify multiple possible paths that lead to churn based on dat…
Summary -Content: Asking various questions about model (algorithm) selection in a Kaggle competition -Conclusion: I now understand the overall picture of model selection. Next, I’ll move on to data understanding! Goal This time, I will a…
Summary - Content: Selecting models (algorithms) for a Kaggle competition with Copilot - Conclusion: Achieved Top 20% in a beginner-friendly competition using an ensemble (hybrid) of CatBoost, LightGBM, and XGBoost! Goal This time as wel…
Summary -Content: Join a Kaggle competition with Copilot and improve the model -Conclusion: It’s possible to generate working code even for elaborate models, but the competition is surprisingly tough Goal This time, together with Copilot…
Summary - Content: Joining a Kaggle competition with Copilot and submitting the smallest possible model - Conclusion: Couldn’t submit on the first try ⇒ Added input/output files and error details ⇒ Submission completed! Goal The goal thi…
