DATA CORRECION

Missing data (cells with no values), and out of range values (for example, a temperature of 1000G in a weather dataset) are listed below on rows 9 and 10.

1. Define "out of range" values only for the cells listed in section 9 For example, you may decide that any temperature t where -30 > t > 120 is out of range. Note: there are no strict rules on this, so decide on some range that looks reasonable
2. Only for cells listed in section 9, replace the missing data and out of range values with values from another row that is most similar to the row where data is missing or out of range. For example:
   
   Row 9 column C (cell 9C) has a value missing. The closest value appears in row 11. Therefore cell 9C is given a value of 77

Row 10 column D has an out of range value. The closest row appears to be row 6, although the city is different. Therefore, cell 10D is given a value of 50. The closest row may also be row 5 as the city if the same, but the temperature difference is greater than for row 6. The similarity algorithm will identify the closest row.

There are a number of similarity algorithms such as Cosine similarity, Jacard similarity etc. You may choose any similarity algorithm. Missing or out of range values will only be numerical (integers or float) values. You may have to use one-hot encoding for text (such as 'Stillwater' for example).

Using the housing data set predict Price using a random regression forest machine learning algorithm. Create a network for both of the following:

• all the features in the dataset
• only the feature with the highest correlation as determined in task 2

Measure the accuracy of both predictions from #5 using RMSE (Root Mean Square Error).

Identify which column or feature has the most impact on the result. For example, if the goal is to predict if a person may get a stroke, the age of the person is the most important variable. It is more important than gender and other variables. Some variables or features or columns such as a person's address is irrelevant when it comes to predicting stroke.

Note: you are required to do a pairwise comparison only. In other words, compare incidence of stroke with one feature only. You will therefore have to do a correlation between each feature (age, gender, etc.) and the output feature stroke. Foe example:

Do a correlation between Age and Stroke, Gender and Stroke, Systolic Pressure and Stroke, ... as Stroke is the output feature.

Create a pipeline for tasks 1 - 4. this is a bonus and is optional.