使用brain.js神经网络比较数组中的字符串和数字

Question

For learning and fun, I wanted to focus on brain.js, but with a different example than it's already described. I have a problem, but probably with a bad approach to this library.

Namely, in a json I have book titles and their prices, while in a train I provide keywords and the price I can give for its occurrence in the title.

my JSON:

  [{ "title": "Bible js", "price": 30 }, 
  { "title": "Bible php", "price": 25 }, 
  { "title": "Bible java", "price": 40 }, 
  { "title": "All about js", "price": 25 }, 
  { "title": "Testing in php", "price": 50 }, 
  { "title": "How to start with java", "price": 40 }]

my code for brain.js:

  const brain = require('brain.js');
  const jsonData = require('./books.json');

  const myMatch = [
    { input: ["js", 40], output: [1] },
    { input: ["php", 20], output: [1] },
    { input: ["How to", 50], output: [1] },
    { input: ["java", 0], output: [0] },
    { input: ["Bible", 0], output: [0] }
  ]

  var config = {
    binaryThresh: 0.5,     // ¯\_(ツ)_/¯
    hiddenLayers: [3],     // array of ints for the sizes of the hidden layers in the network
    activation: 'sigmoid' // Supported activation types ['sigmoid', 'relu', 'leaky-relu', 'tanh']
  }

  var net = new brain.NeuralNetwork(config);
  net.train(myMatch);
  jsonData.forEach(element => {
    let result = net.run(element.title, element.price);
    console.log('--element--');
    console.log(element.title, element.price);
    console.log('--typeof--');
    typeof result;
    console.log('--result--');
    console.log(result);
    console.log('--------------------');
  });

I load book titles and their prices and then I want it to return my degree of interest to given words and the price that interests me or not.

For example, I am looking for books about js for $40 and php for $20. I'm totally not interested in titles with "Bible" or "java". The error I'm struggling with is NaN .

Answer 1

The reason you are getting NaN is you are feeding strings into the net where an array of numbers should go. How many books are you classifying?