Botmistress bot/human classifier

dataanalysis.js

module.exports = function(input) {
  var net = {"layers":[{"mtpd":{},"dptd":{},"msec":{},"dsec":{},"mrt":{},"drt":{}},{"0":{"bias":-3.6984439021091666,"weights":{"mtpd":-6.173024933059706,"dptd":10.997739857335477,"msec":2.7645625300070646,"dsec":2.815832326064508,"mrt":18.0465877829903,"drt":1.155631417743205}},"1":{"bias":2.8315295205493665,"weights":{"mtpd":5.2303592356466035,"dptd":-8.2729870958148,"msec":-2.3778963665239683,"dsec":-2.1532312095623416,"mrt":-14.328315122259214,"drt":-1.0412519208946147}},"2":{"bias":2.715125313888669,"weights":{"mtpd":4.855944732546972,"dptd":-7.991354496280592,"msec":-2.1930872312980267,"dsec":-2.1223653263694544,"mrt":-13.689745670932748,"drt":-0.9648212850166319}}},{"bot":{"bias":-2.2122390536076995,"weights":{"0":-15.7186161148084,"1":11.480607717164341,"2":10.947352405515204}},"human":{"bias":2.1200842597557106,"weights":{"0":15.801011054157996,"1":-11.430970519356372,"2":-10.892672803240483}}}],"outputLookup":true,"inputLookup":true};

  for (var i = 1; i < net.layers.length; i++) {
    var layer = net.layers[i];
    var output = {};
    
    for (var id in layer) {
      var node = layer[id];
      var sum = node.bias;
      
      for (var iid in node.weights) {
        sum += node.weights[iid] * input[iid];
      }
      output[id] = (1 / (1 + Math.exp(-sum)));
    }
    input = output;
  }
  return output;
}

dataprocessing.js

"use strict";

const path = require("path");
const _ = require("underscore");
const ζ = require(path.join(__dirname,"..","zeta.js"));

ζ();

//normalize then average
const mean = x => {
	return _.chain(x)
				.norm()
				.core()
				.value();
};

//normalize then get standard deviation
const dev = x => {
	const m = mean(x);

	return Math.sqrt(
				_.chain(x)
					.norm()
					.map(val => Math.pow(val-m,2))
					.core()
					.value()
	);
};

const nani = x => {
	return _.isNaN(x) ? 0 : x;
};

//function that takes a pulldown array and returns 
const doStuff = raw => {
//obj of arrays of timestamps grouped by date, prolly use for other stuff later
	const datemap = _.chain(raw)
						.reject(val => _.has(val,"retweeted_status"))
						.map(val => new Date(val.created_at).toISOString())
						.groupBy(val => val.slice(0,10))
						.value();

	const rtMap = _.chain(raw)
						.filter(val => _.has(val,"retweeted_status"))
						.map(val => new Date(val.created_at).toISOString())
						.groupBy(val => val.slice(0,10))
						.value();

	const tweetsPerDay = _.chain(datemap)
							.values()
							.map(val => val.length)
							.value();

	const secs = _.chain(datemap)
					.values()
					.flatten()
					.map(val => val.slice(17,19))
					.value();

	const rtsPD = _.chain(rtMap)
						.values()
						.map(val => val.length)
						.value();

	const result = {
		//normed avg of tweets per day
		mtpd: nani(mean(tweetsPerDay)),
		dptd: nani(dev(tweetsPerDay)),

		//" of timestamp second values
		msec: nani(mean(secs)),
		dsec: nani(dev(secs)),

		//" of rts per day
		mrt: nani(mean(rtsPD)),
		drt: nani(dev(rtsPD))
	};

	console.log("inside acctclass:\n",JSON.stringify(result,null,"\t"));

	return result;
};

//hacky bullshit but here's an export that can analyze based on this model
module.exports = pulleddown => {return {input: doStuff(pulleddown) }};

zeta.js

//it me alice yr fav gay tran
//I guess it's as good a time as any to start my "include this fucking everywhere" util file
"use strict";

const _ = require("underscore");

const Zeta = function() {
	_.mixin({
		//given an array of reals, normalize its values to within unity
		norm: arr => {
			const min = _.min(arr);
			const max = _.max(arr);
			return _.map(arr, val => (val-min)/(max-min));
		},

		//simple mean
		core: arr => _.reduce(arr, (m,n) => m+n, 0)/arr.length
	});
};

module.exports = Zeta;