Data Science Methodology

Basically all the steps you should follow to ensure what you are doing makes sense in your data science project.

Following are the steps you have to follow when you are taking up any data science projects.

1. Business Understanding

Understand the question or business requirement . Listen to what is the problem they are facing. For example, A marketing person from a company can come to you and say “ Hey, I want to know which response channel our customer prefer to buy our product.” Thats a legit question he asked. so task is to predict a response channel for each customer. Is a pretty straight forward requirement but it is also important to understand why it is required. Here they could be looking for cost saving one response channel might be costing less than the other so it helps for budgeting. Efficient budget allocation is one reason. Company could be focusing on customizing their marketing campaigns to have more personalized touch and also hit the sweet spot so customer will more likely to buy the product.

All of these reasons adding value to the business and understanding that will help streamline the data analysis process.

2. Analytics Approach:

When we are trying to have an analytical approach, putting the business requirement into a problem statement. Understand what is the type of dependent variable. here what are the options for the response channel for the customers. A customer could buy by visiting the company store , buy a product through online web store. So there are 2 options; let’s call it “Walk-in” and “Web”. So now we have two classes that we want to predict for each user. So in technical term the project just became a Binary Classification Predictive Model.

3. Data Requirement:

Now that we have our dependent/y variable , we need to think of what kind of data/independent variables we will need about each user. This part can include demographic data of a user or user’s web visit history of the web store or it could be walk in store visits that we have on record. All this information can be relevant for our model. Some times we have to gather all the data that is available within the organization’s database and…