kmaher9

package levenshtein

import (
	"unicode/utf8"
)

func computeLevenshteinValue(a, b string) int {
	f := make([]int, utf8.RuneCountInString(b)+1)

	for j := range f {

kmaher9

func enumerateDirectory(location string) []string {
	var files []string
	listing, err := ioutil.ReadDir(location)
	if err != nil {
		panic(err)
	}
	for _, l := range listing {
		files = append(files, location+"/"+l.Name())
	}
	return files

kmaher9

func learn(classifier *bayesian.Classifier) {
	classifier.Learn(businessFiles, Business)
	classifier.Learn(techFiles, Tech)
}

func enumerateClasses() {
	businessDirectory := enumerateDirectory(businessDirectoryLocation) // retrieves a list of all filenames stored in the directory.
	for _, file := range businessDirectory {
		fileContent := readFile(file) // returns the entire contents of the files as a string.
		businessFiles = append(businessFiles, fileContent)

kmaher9

func predict(location string, classifier *bayesian.Classifier) {
	probabilities, _, _ := classifier.ProbScores([]string{readFile(location)})
	fmt.Println(probabilities)
}

kmaher9

func main() {
	classifier := bayesian.NewClassifier(Business, Tech)
	enumerateClasses()
	learn(classifier)
	predict(testFileLocation, classifier)
}

kmaher9

package main

import (
	"fmt"
	"io/ioutil"

	"github.com/jbrukh/bayesian"
)

// exported: the classes that are used to store the learned data.

kmaher9

var fs = require('fs')                           // to get directory listings, and read files - npm install fs
var brain = require('brain.js')                  // contains core functions - npm install brain.js

var businessPath = "path/to/business/dataset"    // location of 100 training files
var techPath = "path/to/tech/dataset"            // second class 100 training files
var testFile = "path/to/test/file.txt"           // file which will be used to test accuracy - belongs to tech

var inputs = []                                  // will be used to store all training data prior to training

var errorThreshold = 0.085                       // the minimum error margin needed to class the nn as trained

kmaher9

var businessFiles = fs.readdirSync(businessPath) // retrieve directory listing
var techFiles = fs.readdirSync(techPath)

for (var i = 0; i < businessFiles.length; i++) {
    var file = businessFiles[i]

    var content = fs.readFileSync(businessFiles + "\\" + file) // opens the file as a stream, reading the whole file into memory
    
    inputs.push({input: content.toString(), output: 'business' }) // the entire file content is stored in a two dimensional array
}

kmaher9

net.train(inputs, { 
    log: true, // prints to the screen the error rate, and current iteration
    errorThresh: 0.05, // a relatively high error margin - but I'm not shooting for 100% accuracy (and it's merely a test)
})

var testFileContent = fs.readFileSync(testFile) // read the entire sample file into memory
var output = net.run(testFileContent) // and then run it! This will predict the class for the testFile
console.log(output)

// output: 

kmaher9

var fs = require('fs')
var brain = require('brain.js')

var bPath = "Z:\\Development\\Javascript\\NodeJS\\Neural Networks\\Brain.JS\\002\\src\\dataset\\business"
var tPath = "Z:\\Development\\Javascript\\NodeJS\\Neural Networks\\Brain.JS\\002\\src\\dataset\\tech"
var sample = "Z:\\Development\\Javascript\\NodeJS\\Neural Networks\\Brain.JS\\002\\src\\dataset\\011.txt"

var net = new brain.recurrent.LSTM()

var bFiles = fs.readdirSync(bPath)