Find the best beneficiary for an accounting firm

Overview

We have just set up our new accounting firm "Albertcounting".

However, as students and young chartered accountants, we need advice.

So we want to find the best person to advise us, so we're creating a profile for him or her. So we're looking for a beneficiary.

Data

Data content

We created our own database for this purpose from 3 sources:

• Directory of chartered accountants: to collect data on beneficiaries
• Papers: to collect data on accounting firms
• LinkedIn: to retrieve their training

We retrieved data from our 3 sources using scrapping methods with the Selenium, Beautiful Soup and Request libraries.

Our aim was really to recover as much data as possible, as there was a real purpose behind it: to contact the beneficiaries. We have a lot of contact information etc.

By joining the 3 datasets, we obtained our final data, in order to find the beneficiary to advise us.

Data Quality/Cleaning/Preparation

We had to deal with data problems. They were not "clean" and had several NaNs. We dealt with these cases.

We also had a typical case to deal with in the 'company size' column. We often found ourselves with data 'between 5 and 10 people'. For this, we set up an algorithm to keep only the numbers in the string (5-10) and average them out to 7.5 employees.

Here is a link to a google drive with the cleaned data : https://drive.google.com/drive/folders/1OU4Gj2SCetQyxOI3BPtzQPSEQ_rVa6zB?usp=sharing

Features

(As a reminder: the "best beneficiary" is the one who offers us the best turnover).

These are the features on which we're going to train our models:

• address
• beneficiary age
• company size
• number of chartered accountants

Machine Learning

Here's how it works:

Search for a type of beneficiary using clustering. Keep the cluster with the best median EBITDA %.

We obtain a group of beneficiaries → best characteristics that favor a good EBITDA percentage

Filter according to our expectations location: Paris region those who have already proved themselves in EBITDA

⇒ a group of ten or so beneficiaries to contact

Bonus: we have a list of these beneficiaries and can contact them.

Installation

Provide step-by-step instructions on how to get a development environment running.

git clone https://github.com/username/projectname.git
cd projectname
pip install -r requirements.txt