Let's imagine you're a talented producer who just released his latest motion picture. Having invested $200 M in the project, you want to predict its box-office revenue to assess the return on investment after its distribution.

‍

The first thing we need is data from similar movies with their box-office and production budgets in order to learn the algorithm. Data acquisition is extracted from Kinopoisk.ru.

We extract data on production budget (production_budget_usd), box-office income (worldwide_gross_usd) and set them in one table in 2 columns:

‍

‍

Now we can graphically display all the results

•⁠  ⁠Where, along X-axis, the maximum value was at $450 million

•⁠  ⁠along Y-axis - the maximum value was at $3 billion

‍

Using Machine Learning techniques and extracted data we can calculate the average profit ratio for every dollar spent.

‍

‍

Thus, we end up with the following approximate forecast: $200 million invested in movie production, brings in roughly $750,000 in box-office success.

‍

Mind you, this model can not be considered accurate, because it's based on 2 variables. But if you want to be provided with more accuracy in the results - add more data set and a few featured variables.

‍

We used this example to demonstrate the Machine Learning Approach. The same technique can be used to calculate ROl and for automation of advertising tools (such as Google Ads and Facebook Business Manager).