AI Exam: 1A: Classify and Prediction

 AI Practitioner - Classify and Prediction Math

Classification Setup

A fisherman named "AI" in a rocky pond named "data" with lots of fish and rocks carries his net that he calls "precision".

Classification (SL) Terms

"Accuracy" = % got right. calcs correctness in environment. % of (fish in net + rocks left in pond) / all. 

Clustering = UL. that auto groups points into segments without needing predefined labels.

Confusion Matrix = outcomes with which you derive accuracy, precision, and recall.  

F1 Score = best fisherman. harmonic mean of fish in your net (precision) and not leaving fish behind (recall). good for binary tasks like churn prediction. F1 = 2 x (Precision x Recall) / (Precision + Recall). best for rocky ponds.

Precision = calcs catch cleanness. High precision means all in net is fish. So is fish in net / all in net. 

Recall = calcs progress. High recall means you didn't miss many fish. fish in net / all fish. 

Prediction Setup

After a while, the fisherman starts guessing the weight of the fish.

Abbreviations

MSE = Mean Squared Error,  RMSE = Root Mean Squared Error

Prediction (SL) Terms

Lift Chart = graphic that shows how much better predictive model vs random guess.

Mean Squared Error = predictions. penalty for big errors. On every error, he has to draw a square on a chalkboard where the sides are the length of error. MSE is the average area of all those squares. measures the avg squared difference between estimated and actual.

Regression = predictions. Think: the guessing game. Ex: The fisherman looks at a fish in the water and guesses, "That fish weighs 5 lbs." Predicts continuous, numerical output values based on input features, mapping relationships to forecast trends like prices, demand, or temperatures.

Root Mean Squared Error = predictions. common metric for regressions. rather than draw squares, he decides just to measure the length. calcs avg difference between predict and actual then square roots it.

R-Squared = scores % variance in prediction explained. 1 is perfect prediction, 0 is worst.