Create a PowerPoint presentation of 10 to 15 slides

Kayode Alaka

Benchmark – Model Building

MIS 690

Model Building

Model building in regression analysis is the process of creating a probabilistic model that best describes the relationship between the dependent and independent variables. The main issues are determining the proper form of the relationship (linear or curvilinear) and deciding which independent variables to include. When developing models, it is often preferable to include both qualitative and quantitative variables. Regression analysis helps to predict the dependent variable on the basis of predictors. The significance of independent variables tells us how much predictors contribute to explain the amount of variation in dependent variable. Businesses can use predictive analytics to identify who their customers and prospects are, how to upsell and cross-sell products and services, and predict consumer behaviour. Building predictive models is an iterative process in which the model is built from a starting point and then improved until it produces a useful organisational outcome.

Methodology approach

Online business trend has been growing very fast in the recent era of life. The success of online business depends on accurate and timely delivery of products. In this report product delivery, accurate deliveries and timely deliveries has been analysed. Logistic regression analysis is used to predict the delay response using the predictors delivery type and delivery provider. Business analyst wants to know delay in delivery is due to delivery type or delivery provider. The variables are coded and description of each code is given below;

Delivery Type ID	Delivery Type Name	Days
1	VIP same day	0
2	Express-1 day	1
3	Express-2 day	2
4	Regular 5-8 days	>=5

Provider ID	Provider Name
1	FEDEX
2	UPS
3	USPS
4	AMAZON DELIVERY

Delay Reason ID	Delay Reason
1	On time
2	Weather Delay
3	Missing item
4	Broken item
5	Wrong item sent
6	Wrong amount sent
7	Mis sort
8	wrong labeling
9	wrong zip code

The required data for the analysis has been stored in csv file that is easily readable in R software. There is not any missing value which need to removed. The data is clean for analysis. The variables customer number, customer ID, shipping date has been removed before analysis.

Modelling Method

For this part of project, I am using Delivery On Time data. I used logistic regression method to see if delivery delay can be predicted using delivery provider and delivery type.

The summary results of logistic regression shows

The results show delivery types and delivery provider does not have strong impact on delay of product delivery. The p value of predictors is above the significant level so we cannot conclude that delay of any item depends on delivery type or delivery provider.

I used to train the data and used 10-fold cross validation to test the model to be better fit. The predicted values of model show that model is good fit to predict the delay response. The null deviance of model is 3993.468 and the degree of freedom is 10408. The lowest null deviance shows that the model is good fit and the deviance value of model is 3987. 4. The deviance value of model included predictors is less than the null deviance. So we can say that model is better fit to predict the delay response.