Programming For Data Analysts Summative Assessment BPP Business School

1.1). Assessment learning outcomes 

This assessment is designed to gauge your understanding, skills and application of common data  analysis techniques used in business and other organisations today. As such you need to  demonstrate your attainment in these areas according to the THREE Module Learning Outcomes  (LOs): 

LO 1: Critically evaluate the principles of programming and apply them in a business context.

LO 2: Critically evaluate the use of code libraries in programming for a business context.

LO 3: Construct a programming solution to solve a defined business problem. 

1.2). Assessment tasks 

This assessment is made up of TWO Parts 

• Part 1 - a coding exercise in data analysis using a Python notebook. 
• Part 2 - writing a business report. 

You will have worked on both these Parts for your Formative Assessment. Now update both Parts for  your Summative Assessment as set out below. You should act on any feedback you received on your  Formative submission, together with your own further learning and development across the module. 

1.3). Scenario 

Zappy Financial Services (ZFS) is a local company that provides small business loans. Last year, loan  applications increased by over 200%, largely because of a concerted online campaign to establish a  strong digital presence. Almost all loan applications and business leads are generated from search engines and digital advertisements, reflecting the decision to increase advertising spend on SEO  channels such as Google, Facebook, LinkedIn and similar platforms. 

Despite a strong digital marketing approach, the current loan application process remains manual.  It requires the online completion of information, including gender, marital status, number of  dependents, education, income etc. Loan decisions are manually categorised as either “approved” or  “rejected.” To date, several of these factors have been considered in the approval decision. All  applications are reviewed and approved by the loan team which, given the recent increase in  volumes, has resulted in skills shortages, longer loan approval times and increased potential  operational and control risk. The current operating model constrains further growth.  

You are employed by ZFS as a Senior Analyst, with programming and data analytics knowledge, as  well as a deep appreciation for the need to balance business growth with a robust control  environment. You will be leading this project with a team of programmers, with various levels of  experience, and have been tasked with providing a scalable solution – that addresses key resourcing  and control risks. 

Specifically, the Board has instructed you to develop and maintain several partial automation  processes that will help the existing loans team, freeing up their time for greater one-on-one  customer contact. You need to provide a data-driven solution while working with a variety of key  stakeholders each with varying objectives such as marketing, internal audit and compliance. You  need to assure the board that you have consulted widely in drawing up requirements and have  considered how the any code produced will be further developed and maintained. You also need to  consider that any code developed will need to be accessible to (used by) other staff such as sales,  customer service and support personnel.

An in-house database administrator (DBA) was able to compile a PDF of past applications which the  loans team are hoping to map to previous loan approval outcomes.

The two files provided by the DBA are:

• A file in PDF format called ‘Loans_Database_Table.pdf’  
• An Excel file, called ‘Zappy Loan Data.xlsx’  

(Note: These same two files are used in the Formative and the Summative). 

The first file has been extracted from business loan records from the previous year, and it includes a  loan approval status field (Yes/No) for each application, allowing the business to map inputs to  outcomes for a possible supervised machine learning exercise.  

The Excel file is maintained by the Sales team and it is currently being saved in a shared folder. This  increases the chance of duplication and missing values.  

You will need to reflect the learnings throughout this module and consider the learning outcomes  particularly LO 3: Construct a programming solution to solve a defined business problem as you  create your answer. 

1.4). Part 1: Construct a Programming Solution 

In Part 1, you will deliver an interactive Python notebook (a . ipynb file) using Google Colab, with the  code used, with comments, to explain the scripts, the libraries used, and the logic. All such  commentary should be written using the built-in markup language (Markdown text) or embedded  using the standard Python syntax. 

The notebook which you create should highlight some of the key findings which you have in the data  and the insights which you can provide to the business. The tasks which need to be completed in the  Python Notebook include the following: 

Task 1: Loan Data Automation 

Create a new .ipynb notebook within Google Colab and load the TWO data files provided by the  DBA. Extract the two datasets from these two files which contains information about past loan  records. The integers shown in each column of the loan datasets are to be interpreted as follows and  you may read them as strings or numbers in your code as you deem appropriate.: 
• Gender: 1-Male, 2-Female 
• Married: 0-Single, 1-Married 
• Dependents: 0, 1, 2, 3+
• Graduate: 0-No, 1-Yes 
• Self_Employed: 0-No, 1-Yes 
• Credit_History: 0-No, 1-Yes 
• Property_Area: 1-Urban, 2-Semiurban, 3-Rural 

You should use Python to load the information of these datasets in memory. You should also add  comments to your notebook, explaining the steps taken to load the data, how you treated the PDF  data, the libraries called and the overall procedure. Recall this will be used for training colleagues in  future. 

Task 2: Descriptive analysis 

First, check the datasets and make sure the data that comes from these two files is valid. Ensure  your loan data is correctly indexed on the LoanID column. 

Then, clean the loan data. Provide an explanation of the steps taken to ensure data preparation for  analysis such as the correction of duplicates, missing values, outliers etc. 

Next, carry out Descriptive analysis on the current loan data. Your notebook file should contain the  code and output of some basic Exploratory Data Analysis (EDA) which must include the following: 

• The total amount loaned by ZFS. 

• The average amount loaned. 

• The average loan term. 

• The total number of applicants broken down into Approved and Rejected, the  number of males and female in each case, and           shown in an appropriate chart.

• The maximum and minimum of amounts loaned and shown in an appropriate chart.

• The number of self-employed who had their loan approved, expressed as a  percentage of all who had their loan approved, and       shown in an appropriate chart. 

Optional (additional marks will be given for the following) 

• The income distribution of all main applicants, showing average and standard  deviation. 

• The top ten applicants by loan amount. 

• The distribution of properties (rural, urban etc) of all loan applicants, shown in an  appropriate chart. 

The code must be commented using the built-in markup facility in Colab or embedded commentary so that it can be maintained in future. The output of calculations and visualisations should be  showed in the notebook and in the report in Part 2. This code and outputs should then be copied  and pasted as Appendix 1 in your Part 2 report.

1.5). Part 2: Report - Business Case  

Using the scenario given, develop a report that proposes and justifies the continued  development and maintenance of the code that has been written by the programming  team. 

The format of the report should include: 

a) Introduction: This should first set out the purpose of the report and its structure. It must  then cover the current business environment of companies like ZFS, the problems that the  programming solution has addressed, and what impact and benefits your proposed  development and maintenance of the code might have on the business. You should also  mention the implications of not proceeding and the kind of human resources needed.  There is no need to discuss project finances, but your introduction must comment on the  following: 

• the source and nature of the supplied dataset and its suitability for data analytics;

• what user requirements have been considered (how will the tool be used); 

• the challenges the team faces in further developing and maintaining the code, the  benefits of using reusable code and any regulatory and ethical implications. 

b) Approach: Describe the approach you took to implement your solution. This should be based on a suitable data analysis lifecycle or framework, such as one covered in the  module, Describe, especially, the key steps taken in loading and preparing the data for  analysis. Use the outputs (calculated values and any charts and visuals) from your EDA in Part 1 to explain the key findings. Key findings can relate to any business insight that might  be gained from the analysis of the historical loan records, as well as any general lessons  about the approach being taken for this project. The ZFS board want to know if your  approach works, is sustainable and worth building on.  

You should provide discussion of 

• The code libraries used (what these are and why they are needed); 

• The reasons for choice of language and platform; 

• The design (e.g., using pseudo code) to aid understanding and allow the design to be  implemented in other languages if needed 

• Testing the code (e.g., by using Excel) to confirm that the programme works; and  • Maintenance of any code to allow controlled and efficient development (e.g., by  version control using Git and good commentary) 

C) Recommendations for future work: Briefly explain how your solution could be further  developed to build a predictive model using the historical data provided. (i.e. A model that can  be trained to predict the outcome of any future loan application.)  
This should include a short explanation of the techniques, libraries, tools, and objective  functions used to evaluate the precision of your recommended predictive model.  

Note: There is no need to implement your ideas in code or build a predictive model 

Your recommendation must also discuss: 
- how existing data collection and data quality may be improved; 
- if other libraries might be used to increase model accuracy; 
- additional training or support the staff will need in order to use these new tools.  - safeguards needed to ensure compliance to        regulatory and ethical standards. 

Conclusions: A brief conclusion summarising the main points in the report.