Deloitte INsight: A Data-driven Industry Analysis Platform

Hi there! My name is Jace Yang. I am one of the core contributor of the Deloitte INsight, an upcoming financial data product that aims to accelerate data-driven industry analysis, especially analysis deep down the subdivision markets.

This repository is to show my work but mainly through pictures and words, as I am not allowed to share any code/source files in public!

Project Website 项目信息

We just launched our product in April!

• Website of the product
• News issued from Deloitte wechat platform

My work 我的工作

Interning for 8 months since the very beginning of the product's R&D period, I have been trusted to take charge of the data division throughout the whole beta stage of this product.

My major contributions here:

• Prototyping the dataflow of the database.
• Building a text classification machine learning baseline.
• Developing a business analytical framework with its data dashboard.
• Designing the web page.

Please bear with me that the language used in most files/charts is Chinese. ^ ^

Data Pipeline

I wrote a SQL-style data processing rundown in R via 100+ tailored function based on dplyr and over 10k+ lines of code. It then helped the team to streamline the workflow and improve the efficiency remarkably.

• The whole dataflow:

  

  • Data cleaning details for 1 datatable as an example, which located 170,000 subsidiary companies' address and industry positioning.

    

• The collaboration workflow

  

Business-Industry Mapping NLP Algorithm

As our team develop a 4-level market segmentation system. My algorithm is to map the company into several niche markets among 1145 different sub-industries.

The way I built the model:

Data Dashboard

I research many existing industry analysis framework, and R&D solutions to empower them by bring our unique data on industry segementation.

To give our user (credit company, security, banker...) a functional tools, I built several visualization demo.

• Demo generated by Highchart:

  

• Demo generated by PowerBi (Link)

• Inside the system: