To train our linear-regression algorithm to generate accurate predictions, we need a cost function. A cost function is one way to analyze how well our prediction function is working for our dataset. It gives us an error rate, which is essentially the difference between the real-world result versus the prediction. The smaller the error, the better predictions we'll make. The cost function takes the real result and the prediction to output the error from our model, like so:
error = result - prediction
There are many different types of cost functions. In our case, we'll use the mean squared error (MSE) cost function, which looks like this:
error = sum((result - prediction)2) / numberOfDataPoints
To make it clearer, we can add ...
