# Arduino number predictor using ML

Hi, can anyone help me how I can make a code that can predict number will be chosen by the user (0-25)? I would like someone to explain me how can I do that from 0 (I never worked with ML before).

What is ML ?

UKHeliBob:
What is ML ?

Machine Learning.

Which is not really relevant on an Arduino forum.

Develop and train a model on your computer in Python or another suitable language. Once it works, port it to Arduino.

If you've never worked with ML before, follow a course, there are some free ones online. Forget about implementing it from scratch on an Arduino for now, get an understanding of the different algorithms first, using existing frameworks.

Pieter

baciuflorin:
Hi, can anyone help me how I can make a code that can predict number will be chosen by the user (0-25)? I would like someone to explain me how can I do that from 0 (I never worked with ML before).

What you want to achieve is not possible. ML prediction works by looking at historical data to infer a trend fitting "future" data.

User chosen number is totally random and can't be predicted.

arduino_new:
User chosen number is totally random and can't be predicted.

A person isn't very good at generating random numbers, so an ML could "predict" the user's next chosen number better than pure chance. (For example, if someone doesn't like to choose even numbers, the ML would eventually "know" not to predict even number either.)

christop:
A person isn't very good at generating random numbers, so an ML could "predict" the user's next chosen number better than pure chance. (For example, if someone doesn't like to choose even numbers, the ML would eventually "know" not to predict even number either.)

Exactly, some "AI" rock-paper-scissors algorithms win 80% of matches against human opponents.

A person "randomly" picking a number is not at all random.

christop:
A person isn't very good at generating random numbers, so an ML could "predict" the user's next chosen number better than pure chance. (For example, if someone doesn't like to choose even numbers, the ML would eventually "know" not to predict even number either.)

The problem is there is zero correlation between historical data and future data. "A person doesn't like even numbers" is very specific (specific criteria is like noise and is removed from the data). ML relies on concrete data pattern in order to "predict".

PieterP:
Exactly, some "AI" rock-paper-scissors algorithms win 80% of matches against human opponents.

A person "randomly" picking a number is not at all random.

ML is a very small subset of AI. This AI may not use any ML models at all and likely use deep learning.