Statistical Analysis

Technical solution

OLS

Ordinary least squares The OLS method corresponds to minimizing the sum of square differences between the observed and predicted values.

OLS_summary

Method Name	OLS_summary
	Method Description	This method will be used to display the OLS summary of the given model.
	Input parameter names	self,dataframe,target
	Input Parameter Description	dataframe: the input dataframe with numeric columns.
		target : target column
	ouptput	OLS summary of the model
	On Exception	Write the exception in the log file. Raise an exception with the appropriate error message

VIF

The variance inflation factor (VIF) is the quotient of the variance in a model with multiple terms by the variance of a model with one term alone. It quantifies the severity of multicollinearity in an ordinary least squares regression analysis.

Correlation: Correlation is a statistical technique that can show whether and how strongly pairs of variables are related

calculate_vif

Method Name	calculate_vif
	Method Description	This method will be used to calculate the vif scores for the cleaned data.
	Input parameter names	self,dataframe
	Input Parameter Description	data: the input dataframe with numeric columns.
	ouptput	VIF values for each of the columns.
	On Exception	Write the exception in the log file. Raise an exception with the appropriate error message

Silhouette Coefficient

Silhouette Coefficient or silhouette score is a metric used to calculate the goodness of a clustering technique. Its value ranges from -1 to 1. 1: Means clusters are well apart from each other and clearly distinguished. 0: Means clusters are indifferent, or we can say that the distance between clusters is not significant. -1: Means clusters are assigned in the wrong way.

Exceptions Scenarios

Step	Exception	Mitigation
Column has mixed values(Integer & number)	Give proper error message	Ask the user to correct the data.
Not all values are numbers	Handle Internally	Convert categorical to numerical values

Seasonality:

Seasonality is a characteristic of a time series in which the data experiences regular and predictable changes that recur every calendar year. Any predictable fluctuation or pattern that recurs or repeats over a one-year period is said to be seasonal.

Stationary Data:

A common assumption in many time series techniques is that the data are stationary. A stationary process has the property that the mean, variance and autocorrelation structure do not change over time