6th

Summary

4-6 sentences summarizing the most important aspects of your model and analysis, such as:

• The training method(s) you used (Convolutional Neural Network, XGBoost) LightGBM
• The most important features
  LGBMRegressor
• The tool(s) you used
  Kaggle Notebooks
• How long it takes to train your model & prediction
  148.3s

Features Selection / Engineering

• What were the most important features?
  Difference from previous day's closing price
• How did you select features?
  After creating the main features such as moving averages and historical volatility, we left the combinations with good evaluations.
• Did you make any important feature transformations?
  No
• Did you find any interesting interactions between features?
  No
• Did you use external data? (if permitted)
  No

Training Method(s)

• What training methods did you use?
  LightGBM
• Did you ensemble the models?
  No

Interesting findings

• What was the most important trick you used?
  I decided to reduce the number of features to one.
• What do you think set you apart from others in the competition?
  I tried something that I thought would be impossible for a high score if I thought about it normally.
• Did you find any interesting relationships in the data that don't fit in the sections above?
  No, it's still hard to read the stock market.

Simple Features and Methods

Many customers are happy to trade off model performance for simplicity. With this in mind:

• Is there a subset of features that would get 90-95% of your final performance? Which features? *
  No

Model Execution Time

Many customers care about how long the winning models take to train and generate predictions:
148.3s

References

Citations to references, websites, blog posts, and external sources of information where appropriate.
Books: Kaggle で勝つデータ分析の技術