Global-Solar-Radiation-in-King-s-Park

Utilizing Machine Learning Regression Model to predict tge global solar radiation using the data that collected in King's Park

First Approach - Trying to predict the average temperature

• Using MySQL and query to handle data
• Using variables including bright sun time, relative humidity and global solar radiation
• Analysing Linear Regression and various degrees of Polynomial Regression's accuracy
• Result: Data Correlation are bad, also may be due to city's heat island effect, it's not that easy to predict with such a little dataset that could be found on the gov db, r2 = 0.47

 

Second Approach - Turn the predicting variable to the GSR

• Turned to use pandas to handle data
• Added 3 more variables including rainfall, uv and mean wind speed
• Preprocessed the data with feature selection and normalization

Result

Polynomial Regression's Degree Optimization

• After the data preprocessing, using degree=6 to optimize the model

 

Prediction Result on GSD with AVG_TEMP, SUN and UV

• r2 = 0.9044

 

Correlation

EMPTY	AVG_TEMP	GSR	RH	SUN	RF	UV	WSPD
AVG_TEMP	1.000000	0.439289	0.294813	0.261222	0.143698	0.360520	-0.069801
GSR	0.439289	1.000000	-0.143867	0.905922	-0.305681	0.122465	-0.100102
RH	0.294813	-0.143867	1.000000	-0.274133	0.219243	0.236309	0.013089
SUN	0.261222	0.905922	-0.274133	1.000000	-0.291873	-0.006021	-0.143479
RF	0.143698	-0.305681	0.219243	-0.291873	1.000000	0.130049	0.064917
UV	0.360520	0.122465	0.236309	-0.006021	0.130049	1.000000	-0.033784
WSPD	-0.069801	-0.100102	0.013089	-0.143479	0.064917	-0.033784	1.000000

Run Locally

1. Clone the project to your repository

git clone https://github.com/Argonaut790/Global-Solar-Radiation-in-King-s-Park.git

2. Change to the project directory

cd .\Global-Solar-Radiation-in-King-s-Park

3. Install the package and dependencies

pip install -r requirements.txt

License

MIT license