[CS116.N11.KHTN] Regression models

1. Members info and picked models:

Name	Student ID	Model
Lê Xuân Tùng	20520347	Lasso Regression
Danh Võ Hồng Phúc	20520275	Ridge Regression
Mai Trung Kiên	20520066	Deep Neural Networks
Cao Văn Hùng	20520193	LightGBM
Lê Phước Vĩnh Linh	20521531	Decision Tree
Nguyễn Tiến Hưng	20520198	Linear SGD
Nguyễn Quốc Huy Hoàng	20520051	Bayesian
Lê Nhật Kha	20520208	Gradient Boosting
Nguyễn Vĩnh Hưng	20520055	Elastic-Net Regression

2. Problem description:

• Weather forecasting is extremely important because it gives information that helps to safeguard human life and property by preventing natural disasters and floods. This is especially crucial for someone like me who is weather-sensitive. Therefore, to better understand how weather forecasting works, and to explore if, based on previous experience observing climate data, can a Machine Learning model that reliably forecasts the weather be built.

• Problem modelling:

  • Input: 3-day continuous weather data for an area (here is Thu Duc City), including temperature, humidity precipitation, pressure, and other variables.
  • Output: The average temperature of the day that need to be predicted.

3. Dataset:

• Name: Thu Duc weather dataset
• Source: crawled from NASA
• Time range: 2000 - 2021 (21 years)
• Attributes: Date, Temperature, Relative Humidity, Specific Humidity, Precipitation, Pressure, Wind Speed, Wind Direction

4. Evaluation method:

• Metrics:
  • MAPE (Mean Absolute Percentage Error)
  • RMSE (Root Mean Square Error)
  • MAE (Mean Absolute Error)
• Protocol: Weather data in 2021 will be used as a test set to give comparison between models. Remain data will be used as a train + validation set. For train and validation, we used K-Fold Cross Validation (with k = 5).

5. Re-implement instruction:

Install essential libs:

pip3 install -r requirements.txt

Note: if you want to run smoothly on MacOS, try to replace tensorflow by tensorflow-macos.

End-to-end run:

sh scripts/train_and_test.sh

Add your model:

First, create your own config file in configs/ folder, using below structure:

name: <Your model name>
args:
  <Your model configuration>
  ...

View the example LinearSGD.yml for better understanding.

Next, create your model class in models/ folder. View example to create it.

Note:

• Your model name must be matched with the class name you created in models
• Your class name must not be the same as scikit-learn model name, this can cause unlimited recursion when initializing.
• ...