# finding 3 variables using regression

## **charge air temperature** using Nordserena (imo 9697026), average signals from every 5 minutes
Those are values from 2 years after selecting only those that are below *100 celcius* (bigger were only at the end and seemed like outliers).

### variables used in regression
* water temperature overboard - **Cenral_CSW_Inlet_temp** (those above 35 celcius were treated as outliers)

* DG power active ( mean over 1st, 2nd, 3rd and 4th)

* Potential temperature at 5 m depth - **pt** (measured only from the beginning of 2018)

* Temperature - **tmax** (also measured only from the beginning og 2018)

In the end we have dataset consisting of 4 variables (5th to be predicted) and 37318 rows. We perform multivariable regression on it and get score of **0.82**.
It means that our predicted values are 82 % better that those we would have by always using mean value.

### formula

**2.02** * Central_CSW_inlet_temp **+ 0.002** * DG **- 1.63** * pt **- 0.37** * tmax **+ 35.6**

### graph