In classification using
PHP-ML we assigned labels to new observation. Regression is almost the same with difference being that the output value is not a class label but a continuous value. It is widely used for predictions and forecasting. PHP-ML supports the following regression algorithms
Regression has the same
predict methods as used in classification.
This is the regression version for SVM(Support Vector Machine).The first step like in classification is to train our model.
// Import library use Phpml\Regression\SVR; use Phpml\SupportVectorMachine\Kernel; // Training data $samples = [, , , , ]; $targets = [3.1, 3.6, 3.8, 4, 4.1]; // Initialize regression engine $regression = new SVR(Kernel::LINEAR); // Train regression engine $regression->train($samples, $targets);
$targets are not class labels as opposed to classification. This is one of the differentiating factor for the two. After training our model with the data we can start with the actual predictions
$regression->predict() // return 4.03
Note that the predictions return a value outside the target.
This algorithm uses
least squares method to approximate solution. The following demonstrates a simple code of training and predicting
// Training data $samples = [, , , , ]; $targets = [3.1, 3.6, 3.8, 4, 4.1]; // Initialize regression engine $regression = new LeastSquares(); // Train engine $regression->train($samples, $targets); // Predict using trained engine $regression->predict(); // return 4.06
PHP-ML also provides with the option of
Multiple Linear Regression. A sample code for the same can be as follows
$samples = [[73676, 1996], [77006, 1998], [10565, 2000], [146088, 1995], [15000, 2001], [65940, 2000], [9300, 2000], [93739, 1996], [153260, 1994], [17764, 2002], [57000, 1998], [15000, 2000]]; $targets = [2000, 2750, 15500, 960, 4400, 8800, 7100, 2550, 1025, 5900, 4600, 4400]; $regression = new LeastSquares(); $regression->train($samples, $targets); $regression->predict([60000, 1996]) // return 4094.82
Multiple Linear Regression is particularly useful when multiple factors or traits identify the outcome.
Now let us take an application of regression in real life scenario.
Suppose you run a very popular website, but the traffic keeps on changing. You want a solution that would predict the number of servers you need to deploy at any given instance of time. Lets assume for the sake that your hosting provider gives you an api to spawn out servers and each server takes 15 minutes to boot. Based on previous data of traffic, and regression you can predict the traffic that would hit your application at any instance of time. Using that knowledge you can start a server 15 minutes before the surge thereby preventing your application from going offline.