1 of 10

Forecast

Prediction

趙若羽

2 of 10

Outline

  • Introduction
  • Dataset
  • Method
  • Result
  • Conclusion

3 of 10

  • Motivation

Studies are dedicated to the prediction of extreme weather through AI tools nowadays. � More closely to the life… the every single day forecast prediction.

​

  • What we want is..

Input a set of data with different features

Output the temperature of next hour

​

  • Background

Decision tree / Linear regression model / Neural network model / …

1

Introduction

4 of 10

  • Our target dataset

The weather data of Leeds, England

​

  • How to collect the dataset

Dataset source:

 

​

Dataset

2

5 of 10

  • Decision Tree Regressor
  • RNN
  • LSTM

​

Ｍethod

３

6 of 10

  • Decision Tree Regressor

Result

4

Decision Tree Regressor

​

​

Features

4

3

Total Depths

9

9

Best Depth

2

2

MSE

1.79

2.39

MSE

MSE

7 of 10

  • RNN

Result

4

Model - RNN

​

​

​

​

Features

4

4

3

3

Epochs

20

50

20

50

MSE

2.081

2.076

1.520

1.510

Loss

Epoch

8 of 10

  • LSTM

Result

4

Model - LSTM

​

​

​

​

Features

4

4

3

3

Epochs

20

50

20

50

MSE

1.825

1.821

1.540

1.511

Loss

Epoch

9 of 10

  • LSTM

Conclusion

5

10 of 10

Thanks