# 5th solution more data+FE+ensemble

Thank you to the organizers for organizing this competition, and thank you to all the participants for their efforts. I have learned a lot from this competition.
[Here](https://www.kaggle.com/code/yunsuxiaozi/brist1d-top5-solution)
is my solution, with a score of 2.4067 on the public leaderboard.Below, I will briefly explain my solution.
Problem conversion (obtaining more data)
Although the competition requires us to predict the next hour's data based on 6 hours of data, analyzing the correlation between the target and time reveals that the only time that truly correlates with the target is the 20 minutes closest to it. Data that is distant in time may become noise. So here we have transformed the problem into 'predicting the next hour's data based on the data from the previous x minutes'. After the final experiment, it was found that only about 45 minutes are enough.
I have also constructed the test data here, and I am not sure if the competition allows it, but I can guarantee that performing this operation on the training data will definitely improve the performance of the model. On the one hand, it removes noise, and on the other hand, it increases the amount of data.
2.Feature Engineer
A lot of feature engineering has been done here, such as the common sine cosine transform for temporal features, constructing cross features (addition, subtraction, multiplication, and division) in the original features, and calculating the temporal diff features.
3.ensemble
In addition to using three models myself during model training, I also referenced open-source code from others.Here is the code.
submission=pd.read_csv()

submission2=pd.read_csv()

submission3=pd.read_csv()
submission[]=*submission[]+*submission2[]+*submission3[]
submission.to_csv(,index=)
submission.head()
4.Yunbase
Welcome to use
[Yunbase](https://www.kaggle.com/code/yunsuxiaozi/yunbase)
in other competitions.