Important_notes.txt

2.	In your opinion, what are the limitations of working with the data in 10-minute intervals directly? – 2 missing days + 6 days with incomplete 10-minute interval readings.
3.	Final term project dataset: its 10-minute data. Do you think aggregation is required? Aggregation to days leads to 4500 rows. If this is acceptable then how do you suggest I take an aggregation – an average of the entire day?
4.	While aggregating - Max temperatures, Min temperatures, and Average temperature of a day as separate time series to predict the next day’s max, min, and average?
5.	Can I use a different variable from the dataset for the univariate time series analysis and a different variable for the linear regression model?

-	Have to take average time series for unbiased model – can do min and max as well.
-	Feature reduction using random forest or PCA or condition number
-	Drift method for data imputation. 
- Linear regression - OLS from statsmodel can be applied for time series as well with shuffle false. The only issue is that we cannot make h step prediction. Only single-step prediction can be made.