Article directory
Related Reading:
- [Machine Learning Competition + Notes] Industrial Steam Volume Forecast: Comprehension of Competition Questions (1)
- [Machine Learning Competition + Notes] Industrial Steam Volume Forecast: Data Exploration (Part 2)
- [Machine Learning Competition + Notes] Industrial Steam Volume Prediction: Feature Engineering (Part 3)
- [Machine Learning Competition + Notes] Industrial Steam Volume Forecast: Model Training (4)
- [Machine Learning Competition + Notes] Industrial Steam Volume Forecast: Model Verification (V)
- [Machine Learning Competition + Notes] Industrial Steam Volume Prediction: Feature Optimization (6)
- [Machine Learning Competition + Notes] Industrial Steam Volume Forecast: Model Fusion (Part 7)
Competition address:Industrial Steam Volume Forecast_Learning Competition_Tianchi Competition
1 Comprehension of competition questions
1.3 Data Overview
1.3.2 Data description
A total of 38 fields from V0 to V37 are feature variables, and the target field is the target variable.
1.4 Evaluation indicators
The prediction results are based on the mean squared error MSE (Mean Squared Error) as the evaluation criterion. The calculation formula is as follows: M S E = S S E n = 1 n ∑ i = 1 n w i ( y i − y i ^ ) 2 MSE=\frac{SSE}{n}=\frac{1}{n }\sum_{i=1}^{n}{w_i(y_i-\hat{y_i})^2} MSE=nSSE=n1i=1∑nIni(yi−andi^)2
You can directly call the function to calculate MSE in sklearn:
from sklearn.metrics import mean_squared_error
mean_squared_error(y_test,y_predict)
1.5 Competition question model
1.5.3 Problem-solving ideas
In this competition question, it is necessary to predict the value of steam volume based on a total of 38 characteristic variables provided from V0 to V37. The predicted value is a continuous numerical variable, so this problem is solved by regression prediction.
The algorithms used in regression prediction models include linear regression (Linear Regression), ridge regression (Ridge Regression), LASSO (Least Absolute Shrinkage and Selection Operator) regression, decision tree regression (Decision Tree Regression), Gradient Boosting Decision Tree Regression.