COMP4331 Assignment 1 Question 2

README

1. Use Google Chrome 61.0 or above.
2. Navigate to this page.
3. Press F12.
4. Choose the "Console" tab.
5. Paste in the script and press enter.

question2.js

var labels = ["name", "height(cm)", "weight(kg)" ,"eye-color", "hair-color", "region"];
var dataset = [
	["Abel","180","73","hazel","black","Europe"],
	["Bob","183","75","brown","brown","Europe"],
	["Carl","176","70","hazel","blond","Europe"],
	["Dale","168","60","brown","black","Asia"],
	["Eric","174","65","brown","brown","Asia"],
	["Felix","170","59","brown","blond","Asia"],
	["George","180","78","blue","black","America"],
	["Howard","179","80","hazel","brown","America"],
	["Igor","175","75","blue","blond","America"]
];

// Question 2a
// Calculate the prior probabilities for each class
console.log('Quesiton 2(a): Calculate the prior probabilities for each class');
var q2a = dataset.map((row) => {
  // Get region
  return row[5]
}).reduce((arr, region) => {
  // Reduce to region and count
  if (!arr[region]) arr[region] = 0;
  arr[region] += 1;
  return arr;
}, {});
// Output table of region and probability
console.table(Object.keys(q2a).map((region) => {
    return [region, q2a[region] / dataset.length];
}));

// Question 2b
// for each class C_i and test sample in part 1(h), the value of P(sample|C_i);
console.log('Question 2(b):');
console.log('Output for each class C_i and test sample in part 1h, the value of P(sample|C_i);');
var samples = [
	["Jack","173","65","brown","black","?"],
	["Todd","184","75","blue","blond","?"]
];
var regions = [...new Set(dataset.map((row) => row[5]))];
var average = (arr) => arr.reduce((a, b) => a + b, 0) / arr.length;
var stddev = (arr) => {
	let avg = average(arr);
	return Math.sqrt( arr.reduce((a, b) => a + Math.pow(b - avg, 2), 0) / arr.length );
}
// Calculate probability given continuous attribute A_k
var continuous_p = (x_k, values) => {
	let stddev_values = stddev(values);
	let avg_values = average(values);
	let exponent = Math.pow(x_k - avg_values, 2) * -1;
	exponent = exponent / 2 / stddev_values / stddev_values;
	return 1 / ( Math.sqrt(2 * Math.PI) * stddev_values ) * Math.pow(Math.E, exponent)
};

let answer = [];
samples.forEach((sample) => {
	let name = sample[0];
	console.log('For ' + name + ':');
	regions.map((region) => {
		let rows = dataset.filter((row) => row[5] === region);
		let working = [];
		working.push([
			'P(' + labels[1] + '=' + sample[1] + '|' + region + ')',
			continuous_p(+sample[1], rows.map((row) => +row[1]))
		]);
		working.push([
			'P(' + labels[2] + '=' + sample[2] + '|' + region + ')',
			continuous_p(+sample[2], rows.map((row) => +row[2]))
		]);
		// Calculate probability for A_k with discrete values with Laplace correction
		// For simplicity of code, hardcoded to 3 for number of unique values
		working.push([
			'P(' + labels[3] + '=' + sample[3] + '|' + region + ')',
			(rows.filter((row) => row[3] === sample[3]).length + 1) / (rows.length + 3)
		]);
		working.push([
			'P(' + labels[4] + '=' + sample[4] + '|' + region + ')',
			(rows.filter((row) => row[4] === sample[4]).length + 1) / (rows.length + 3)
		]);

		// Calculate the probability of P(sample|C_i)
		let probability = working.map((row) => row[1]).reduce(function(a,b){ return a * b; });

		// Get answer for Part 2(c)
		// For simplicity of code, quick and dirty cut-off at 1e-4 instead of finding the largest value
		if (probability > 0.0001) {
			answer.push([name, region, probability]);
		}

		// Output working and answer for 2(b)
		console.table(working);
		console.log('P(' + name + '|' + region + ') = ' + "\r\n" +
			    working.map((row) => row[0]).join(' x ') + "\r\n" +
			    probability);
	});
});

console.log('Question 2(c): Predicted Region for each test sample in part 1h:')
console.table(answer);