Overview / Usage

A lot of Investment companies and Venture Capital firms tend to do due-diligence on various businesses (startups) before actually taking a decision to invest in the business. This basically means that the investment company goes through the profits, money spent on R&D, the state the company is in etc. Then based on how the startup spends money in these areas can decide if they even want to consider investing in the business. It serves to be part of the scrutinizing process of any investment agency in reducing turnaround times and focusing on ideas that matter the most.

This is solving several problems that VC Companies, Angel Investors, and also for accelerators and incubators. This will reduce the administrative procedures when scrutinizing applications of startups. Only once they pass this phase can they actually go out and actually consider taking them in. This creates a layer of the application process but it is almost instantaneous feedback on whether the startup passes or fails and goes into the next stage of selection.

For now, this is the core of the algorithm which needs more data to be fed in, and other variables can be taken into consideration. The biggest thing of all the above investment companies or incubators look for is how much profit a business can actually make in the future. So this model basically predicts how much profit a business would potentially make based on how much they spend on R&D, Administration, and Marketing.

This is still under progress, and alot can be done with this algorithm such as include more variables other than marketing, R&D and Marketing. Also can include graphs for comparison. And build on top of this, but this is the core of the model.

Methodology / Approach

A dataset of 50 startups and how much they spent on marketing, R&D, and administration, as well as the state and profit they make, was used to train the model.

I normally like to develop on Spyder which is an Anaconda tool for Data Science development. It has a simple way to track data in the variable explorer.

A simple Multi-linear algorithm was used to build the model from the sci-kit learn the framework. I also used Intel's DAAL for python. Additionally optimized the algorithm further using backward elimination.

Technologies Used

Intel distribution for Python, matplotlib, numpy, sklearn, pandas
Entire structure on Anaconda environment

Repository

https://github.com/akhilboddu/VentureCapitalTool_algorithm