HSX Forum

Movies

Reply Topic List All Forums

Soooo wanted to let this go, but I really hate ignorance

Posted by: AB_Excello on Jun 06, 23:56 in response to horlicks' post Its called math...

You still don't have the slightest clue about the term ACCURACY. Your grade school math is all well and good, but you have no clue how to calculate for accuracy in a predictive model...which guess what.....that's what you're defending.

That's why you have to define what constitutes accurate in your system. Let's just say within 75% by your fuzzy math/fuzzy logic is within your definition of accuracy for your system.

Within 75%+/- = 31

Not within 75% = 22

To get accuracy you take all cases that fit into the parameters and divide by that number and the number that didn't make the parameters. Actual formula is (Total positives + Total negative)/(Total positives + Total negatives + False positives + False negatives) but have to put it as simplistic as I can obviously.

And your accuracy is 58.5% at 75%

Or 49.05% at 80% 26/53

Hence why I keep asking what the number is you would deem ACCURATE....but that was presumptive of me because I presumed you knew what accurate meant....my apologies....gave you too much credit.

And you should get the idea after that. Don't know how you get 1-error as accuracy. HSX thing or not, doesn't matter in this case because you are making a case for a PREDICTIVE MODEL. So you have to use.....a predictive formula for accuracy.

You can try to bury someone all you want, fact is I don't have the data that you have and I don't want it and find no use for it. So wow, you got me....your total tweets were from Monday to Thursday....showed me.

Is there value in data from Twitter.....possible....same possible I stated in my initial post. But the stuff you still overlook is you have to use a specific value to age demos/ethnic demos/et all because they aren't accounted for fully in Tweets. The no PI reference...though now I'm starting to doubt you know Pi in anything other than Apple or Blueberry form.

But I'll leave it at you didn't know how to calculate accuracy and if you read any of this post and still feel like arguing about its definition....well, your special ed teachers did a helluva job getting you up to the point where you can use the computer and Tweet. They should get awards.

So, the Twitter prediction 'myth' scored another win this weekend. horlicks Jun 06, 12:01

Not bad at all. Thanks for your time in compiling the info. {nm} grammar Jun 06, 12:04

Impressive average but I still don't understand what it's counting. {nm} BuenaOnda Jun 06, 12:43

Yah, I don't see how the tweet count and (theater?) count are calculated into a prediction. And, don't see what that ratio is. {nm} edzep Jun 06, 12:52

Trade secrets, I bet :) {nm} RogerMore Jun 06, 13:36

It's explained pretty well on the site. {nm} secretstalker Jun 06, 15:51

Fine. But, if the OP was meant to persuade us dunderheads, then it should have been explained. As is, it's kind of meaningless. {nm} edzep Jun 06, 16:12

Sure, here's a bit more info. horlicks Jun 06, 18:50

Helpful. You're certainly putting a lot of work into it. {nm} edzep Jun 07, 04:10

Since this is probably aimed at me, I'll give you a couple of good replies AB_Excello Jun 06, 14:20

If 74% is a D+, there isn't a box office predictor who does better than a C- {nm} RogerMore Jun 06, 14:37

And that's fine AB_Excello Jun 06, 15:07