linear-regression-model

An npm package to make it easier to deal with a handful of values, and try to model them in one of the most used mathematical models, with an R-like accuracy algorithm

Setup
Methods
Contribution Rules

Setup

Requirements:

Node.js installed
The ideal version, to run the package, is LTS(16.13 by the time it is being published),
however, older versions shouldn't have any issues, as the package does not use any
other npm packages, or fancy, new methods, not supported by older versions
Installing the module

npm install linear-regression-model

Importing the module

First, there is the Linear Regression model between two different datasets, in relation to each other
Which will look like something similar to this:

import {LinearModel} from 'linear-regression-model'
// or
const {LinearModel} = require("linear-regression-model");
// if it is preferred not to use destructuring, or constants use
// var LinearModel = require("linear-regression-model-model").LinearModel;

Or if the use case for this is pending more to the behavior of one single dataset
overtime, this will be more fitting:

import {LinearModelOverTime} from 'linear-regression-model'
// or
const {LinearModelOverTime} = require("linear-regression-model");
// if it is preferred not to use destructuring, or constants use
// var LinearModel = require("linear-regression-model").LinearModelOverTime;

And to use the Correlation class, it will be some like this:

 import {Correlation} from 'linear-regression-model'
 // or
 const {Correlation} = require("linear-regression-model");
 // if it is preferred not to use destructuring, or constants use
 // var Correlation = require("linear-regression-model").Correlation;

Although if your use case is the overtime behaviour, i would advise the use
of the method in the LinearModelOverTime class
But if you are using the LinearModel class, it is really up to you

Instantiate the classes

const lm = new LinearModel([1, 2, 3, 2, 3, 4], [3, 4, 5, 4, 5, 6]);
// mandatory to pass two, same sized, all number, arrays, being the
// orientation (x, y), being x the independent variable and y the
// dependent: y changes according to how x changes, basically

const lm = new LinearModelOverTime([1, 2, 3, 2, 3, 4, 3, 4, 5, 4, 5, 6]);
// mandatory to pass one, all number, array, being the orientation y, the
//dependent variable: y changes according to how x changes, the later is
// generated automatically for a better representation of the behavior overtime

 const corr = new Correlation([1, 2, 3, 2, 3, 4], [3, 4, 5, 4, 5, 6]);
// mandatory to pass two, same sized, all number, arrays, being the
// orientation (x, y), being x the independent variable and y the
// dependent: y changes according to how x changes, basically

Let it be clear, it is necessary to pass all number arrays, with more than 1 value,
in order for the algorithm to work properly, it will try to convert all the elements
to numbers, but if that is not possible, the code will crash.

Methods

Both the classes have similar methods, the biggest difference between the classes being
if the X axis dataset is informed or generated, therefore, the method will be showed using the LinearModel class, the use for the LinearModelOverTime class is the exact same though

Just to make it clear, it is necessary to pass the parameters as said in instatiating the class, above

LinearModel and LinearModelOverTime

getDataset

 lm.getDataset()
 // returns the dataset Informed in the Y axis

getXAxisValues

 lm.getXAxisValues()
 // returns the dataset in the X axis, informed previously or not

(static) radsToDegs

 LinearModel.radsToDegs(rad)
 // converts an angle in radians to degrees
 // returns the angle in degrees

(static) getMean

 LinearModel.getMean(dataset)
 // Utility method to calculate the mean of a dataset

(static) getMode

 LinearModel.getMode(dataset)
 // Utility method to calculate the mode of a dataset

(static) getMedian

 LinearModel.getMedian(dataset)
 // Utility method to calculate the median of a dataset

getDatasetLength

 lm.getDatasetLength()
 // returns the dataset length

getSumOfDatasetValues

 lm.getSumOfDatasetValues()
 // returns the sum of all the elements in the dataset on y azis

getSumOfXValues

 lm.getSumOfXValues()
 // returns the sum of all the elements in the dataset on x azis

getSlope

 lm.getSlope()
 // returns the slope of the equation: the "m" on the y = mx + n function
 // which is the tangent of the inclination angle

getAngleInRadians

 lm.getAngleInRadians()
 // returns the inclination angle in radians, like 3/4 meaning  3/4 rad

getAngleInDegrees

 lm.getAngleInDegrees()
 // returns the inclination angle in degrees, like 43°,
 // for example 3/4 rad ~~ 43°

getDatasetBehavior

 lm.getDatasetBehavior()
 // returns the overall macro behaviour of the dataset

getDatasetBehavioralIntensity

 lm.getDatasetBehavioralIntensity()
 // returns the intensity of the behaviour previously measured

getLinearCoefficient

 lm.getDatasetBehavioralIntensity()
 // returns the intensity of the behaviour previously measured

getLinearCoefficient

 lm.getLinearCoefficient()
 // returns the linear coefficient of the equation: the "n"
 //on the y = mx + n function

getCoefficients

 lm.getCoefficients()
 // returns both the linear coefficient and the slope
 //of the equation: the "m" and the "n" on y = mx + n

getLinearEquation

 lm.getLinearEquation()
 // returns both the function as a string to be displayed
 // and an actual js function to make predictions, for example
 // function(x): returns Y, using the same method as the dataset

getR2

 lm.getR2()
 // returns the coefficient of determination(R²) to find
 // the accuracy of the linear regression just calculated

Correlation

getCorrelation

 lm.getCorrelation()
 // returns the correlation between the datasets

getCorrelationInterpretation

 lm.getCorrelationInterpretation()
 // returns the interpretation of the correlation index
 // between the datasets

Correlation (class)

(static) getMean

 Correlation.getMean(dataset)
 // returns the mean of the dataset

(static) getDifferenceFromMeanAndElements

 Correlation.getDifferenceFromMeanAndElements(dataset)
 // returns the difference between the mean and the elements
 // of the dataset, so can be used to calculate variance
 // or standard deviation, for example

getCorrelation

 corr.getCorrelation()
 // returns the correlation between the datasets

getCorrelationWay

 corr.getCorrelationWay()
 // returns the way/sign the two datasets are correlated
 // to each other

getCorrelationIntensity

 corr.getCorrelationIntensity()
 // returns the intensity of the correlation between the datasets

getCorrelationInterpretation

 corr.getCorrelationInterpretation()
 // returns the interpretation of the correlation index
 // between the datasets

Contributing

Well, since this is a really simple package, contributing is always welcome, just as much as creating issues experienced with the package

In order to better organize this contributions, it would be ideal that all PRs follow the template:

PR Template

WHAT:
A brief description of the improvements

WHY:
A explanation on why those changes were needed, necessary, or at least, why is was on the best interest of the package users

CHANGES:
List of changes made, can be the name of the commits made, or a simple changes list

Commits

Ideally the commits should make use of the convention of Conventional Commits
Something i recommend is the usage of either the Commitizen terminal extension or the Commit Message Editor VSCode Extension

nikolas-virionis/linear-regression-model