Trevor Huang - ML Project pt 2

[trevorhuang1]

ML Project pt 2 (1/2)

• I worked with Matthew on this project
• Theme: we both worked on the same cooking game last trimester, so we decided to make modifications to the shop feature
  • Let 'Em Cook is a fun cooking game which rewards players every time they bake something. Matthew and I noticed that users could just keep farming points indefinitely, since low level items could be made at a faster rate than the cost of the items. To counter this and incorporate corporate greed, the suppliers are beginning to inflate their prices!
  • As time increases, the model predicts what future inflation rates will look like based on historical CPI values, so it increases the prices accordingly.

Next Steps

• Implement supply and demand --> the more people who buy the product, the more inflated its price
• Make inflation more controlled --> perhaps making it only 10 secs/day instead of 10 secs/month because the prices get out of hand extremely fast
• More data about outside events in the world which could help accuracy. For example, the model does not predict based on real world events such as pandemics, financial depressions, and even news sources

We used a custom dataset which tracks United States CPI (customer price index

Frontend to backend api

• API code
• Frontend fetch request
• HTTP requests (it is returning 200 and the date being used to predict inflation prices is printed out)

Applying to a team project

• This is what the project looked like without inflation. The shop allows users to access both default and paid items, and inflation in this game will be used to blunt progression a bit.
• Inflation within the game (ever 10 seconds is 1 month within the game --> this is sped up just to demonstrate the effect of inflation on prices)
  • Essentially a post request is sent to the backend which has a date in the body. Every 10 seconds, the body is updated to be 30 days later

Adding as supplemental part of your project

• Machine learning wasn't its own feature, but integrated into a previous feature we made

Frontend with purpose

• The game is much more realistic to the modern 21st century now
• Users can no longer continuously farm points and need to plan strategically to maximize their points accounting for inflation.
• Inflation being printed out below

[trevorhuang1]

ML Project pt 2 (2/2)

• Here are some screenshots of the actual code

Frontend

• Function which updates the price of the items and adjusts the HTML accordingly

• The date in the body being sent in the fetch request is also being converted to a date, changed to a month later, and then converted back to the string format it was in the beginning

• After the api returns the predicted inflation rate, the frontend prices are multiplied by this inflation rate to obtain the new prices. This is being done every 10s (this can be changed later)

Backend

• Cleaning up the data. There are columns for CPI the historical date associated with it. We also calculate the inflation rate based on the percentage change of the CPI.

• Linear regression. We used linear regression and not logistic regression in the titanic code because the data is not catagorical

• Predict future inflation rate with the linear regression

Demo

• Prices after first inflation

• Initial amount of points

• Points after buying an inflated product (yeast for 22.35)

• New point total