So this is a overview of what I did in this Problem Statement-->

### Data Preprocessing
Though the nan values didnt seem to have a lot effect on the corresponding radiation values, I tried to impute 'em with mean values through SImpleImputer module.
Initially there was an issue with imputed due to the Date time column not being numerical, but that was sorted out when I considered the other parameters ignoring the date time column for imputation(as it really doesnt make a difference).

Data Correlation and Visualizing
Correlation and visulaizing using heatmap, bar graph and other plots gave a pretty fair idea of which parameter has how much effect on our target variable aka Radiation, where maximum correlation was observed with Temperature, the value being '0.73'.
And the least correlation was seen with Wind Direction being '-0.23'.

Decision Tree Regressor Model
Data was split into training and testing models,
Test size = 0.2
Random State = 42

outcome-->
Mean Absolute Error: 6.597076794310784e-13
Mean Squared Error: 6.749341995782093e-25
R-squared: 1.0
This R2 score was very impressive still another model was to be tried out.

Bayesian Linear Regression Model
As the temperature correlation with the Radiation(target variable) was seen to be the highest(0.73), polynomial comparison of both were tried out in this model where i tried the outcomes and uncertainties achieved with different degree values

Again the inputs were-->
Test size = 0.2
Random state = 42

Outcomes:

Degree = 1
Mean Absolute Error: 6.597076794310784e-13
Mean Squared Error: 0.002805110928405552
R-squared: 0.051305786955890054

Degree = 2
Mean Absolute Error: 6.597076794310784e-13
Mean Squared Error: 0.002779660710633128
R-squared: 0.059913102294800735

Degree = 3
Mean Absolute Error: 6.597076794310784e-13
Mean Squared Error: 0.0027780267839264423
R-squared: 0.060465699625435354

Degree = 4
Mean Absolute Error: 6.597076794310784e-13
Mean Squared Error: 0.0027771394433982097
R-squared: 0.06076579999423337

Degree = 5
Mean Absolute Error: 6.597076794310784e-13
Mean Squared Error: 0.0027761076925069987
R-squared: 0.061114740241081766

The mean absolute error did not seem to differ whereas the Mean squared error and the r2 score improved with icreasing degree.
It was thus observed that the R2 score for this model seemed to be improving with increased degree of the polynomial parameter, which i couldnt interpret why.

Radiation is propotional to Temp^4 (By Stefan's Formula)
so the r2 score should have been best with degree=4 but still the r2score improved with degree=5

Overall, Mean absolute error did not seem to make a difference in both models but The mean squared error was much less in the bayesian linear regression......
Due to a favourable R2 score obtained in the Decision tree regressor I will be going with the Decision Tree Regressor .