KPMG_Virtual_Internship_Challenge

Customers and transactions analysis

Objective: This challenge consists of 3 modules and is provided by KPMG through the Virtual Internship Program. The 3 modules are:

1. Data Quality Assessment: Perform EDA using Python and build visualization dashboards using Power BI. Tasks to be performed in this module include:

   • Check data quality:

   • * Check data accuracy: if values are close to true values.
     * Check data completeness: fill nan values with values that make sense. Drop rows if neccessary.
     * Check data consistency: check if values contradict to normally observed trends.
     * Check data currency: check if data are up to date.
     * Check data relevancy: drop columns or values that do not make any differences in the analysis or ML model development.
     * Validate data: check if data are within allowable ranges of values, and check for outliers. Ex: someone's age cannot be over 200.
     * Check for duplicates: remove duplicates to avoid double dipping in the ML model.         
     

   • Data transformation/ treatment

   • Understand data distribution and correlations between features

   • Dashboard visualizations

2. Data Insights: Predict customer trends and behavior using supervised Machine Learning algorithm, recommend which of the 1000 new customers should be targeted to drive the most value for the organisation. Tasks to be performed in this module include:

   • Feature engineering

   • Models development and comparison

   • Best model selection and hyperparameter tuning

   • Use model to predict which of the new 1000 customers should be targeted.

3. Data Insights and Presentation: Summary findings in 3 dashboards and a presentation, while answering important questions such as

   • What are the trends in the underlying data?

   • Which of the customer segment has the highest customer value?

   • What is your proposed marketing and market growth strategy?

   • What additional external datasets maybe useful to obtain greater insights into customer preferences and propensity to purchase the products?

Data Type: Structured data (Nemerical and Categorical), time-series. Since it's a structured dataset, the clean dataset can be stored in a SQL database such as SQL Elephant or PostgreSQL.

Data Source: Dataset was provided by KPMN's client, Sprocket Central Pty Ltd. (Provided in the Resources folder of this repo)

Datasets:

 * Customer Demographic

 * Customer Addresses
 
 * Transactions data in the past 3 months

The datasets are related to each other by the customer_id column as shown in the diagram below:

*Results Summary: (In progress)

1. MODULE 1: (Complete)

• A list of recommendations for Data Treatment is provided in the Documentation folder as well as the Module 1 notebook in the Results folder.

• 2 dashboards were created using Power BI. The file is also included in the Results folder and it can be published for sharing. Snapshots of the dashboards are shown below:

  

2. MODULE 2: (In-progress)