Why Prediction Possible with Data || Lesson 1 || Machine Learning ||

Описание: Why Prediction Possible with Data
Welcome everyone! Today marks our first session in Machine Learning. Before we dive into algorithms, we must address a fundamental question: Why is prediction possible with data? We take an example and understand the logic behind it.

Let’s ground our theory in a real-world scenario. Imagine a server starting her very first day at a restaurant. To understand her potential earnings, she decides to record every transaction. By tracking the relationship between the Bill Amount and the Tip Received, she is creating her first dataset. This ledger is more than just numbers—it is the raw material from which we will extract predictive patterns.



To make a good prediction, one day of data isn't enough. We need to see if the pattern repeats. So, our server tracked her second day as well. Let’s take a look at the dataset.



Now that we have gathered data from different days, we can combine them into one single view. This is our complete dataset. By looking at all six observations together, we move from small stories to a statistical reality. This consolidated table is exactly what a machine learning model uses to find the ultimate pattern.
Because our dataset is still relatively small, we can use our natural human intuition to spot patterns. By simply observing the numbers, we can see clear trends emerging. For instance, notice that when the bill amount is near 500, the tip tends to hover around 50. However, as the bill increases to nearly 800, the average tip shifts upward to about 65. This mental mapping we are doing—finding a relationship between input and output—is the core logic that a machine learning model automates at a much larger scale.


Now that we have successfully identified the repeating patterns in our data, we can apply those insights to make real-world predictions. For instance, if a new customer arrives with a bill amount of 500, what would the predicted tip be? By looking back at our historical patterns, we can see that a bill of this size consistently results in a tip of 50. This is the essence of predictive modelling.