ZhaoyiW / Solar-Radiation-Linear-Regression

Solar radiation prediction

Solar-Radiation-Linear-Regression

Backgroud and Goal

Solar radiation measurement is hard and costs much. The goal of this project is to build a regression model to predict the solar radiation as accurate as possible.

Modules

• pandas: data processing
• numoy: linear algebra
• seaborn: data visualization
• matplotlib: data visualization
• datetime: manipulate date time types of data
• sklearn: maching learning

File Descriptions

SolarPrediction.csv

• The original dataset
• Kaggle was the last destination in the provenance of the data
• Original source: NASA
• Variables captured within the dataset are solar radiation, temperature, humidity, barometric pressure, wind direction, wind speed, and sunrise/sunset based on Hawaii time.

DataWrangling.ipynb

• A Jupyter Notebook to clean and wrangle the data
• A look at the data: data types, distributions of numerical fields
• Data cleaning
• Feature engineering: add dummy variables and higher-order terms
• Correlation matrix

Solar_Features.csv

• The output file from DataWrangling.ipynb

LinearRegression.ipynb

• A Jupyter Notebook to build linear regression models
• Model 1: Linear Regression (No Higher-Order Terms)
• Model 2: Linear Regression (With Higher-Order Terms)
• Model 3: Ridge Regression
• Model 4: Lasso Regression

Best Result

R squared = 0.7248

License

This project is under MIT License.

Author

Zhaoyi Wang