Predict future values of time-based dataset in PHP

Question

For each year the given dataset contains the average price for the item sold, take this for example:

╔══════╦═══════════════╗
║ Year ║ Cost of flerg ║
╠══════╬═══════════════╣
║ 2007 ║            13 ║
║ 2008 ║             7 ║
║ 2009 ║             8 ║
║ 2010 ║             9 ║
║ 2011 ║            12 ║
║ 2012 ║            13 ║
║ 2013 ║             9 ║
║ 2014 ║            11 ║
║ 2015 ║            14 ║
╚══════╩═══════════════╝

Now I want to estimate how much it'll be worth in 2016, 2017, ...

I have looked at some linear regression functions, but they only produce a slope and intercept number for a graph.

So does anyone know of a way to go about this?

You can't predice the future, almost without a crystal ball. That @Sougata means makes sense for me. Good luck. — Marcos Pérez Gude, Nov 20 '15 at 08:44

score 0 · Answer 1 · answered Nov 20 '15 at 08:54

You could try to fit a line through the points using the least squares method, which assumes you know the approximate law of the dataset distribution.

You could also try training a small perceptron or a neural network (depending on whether you think it is a linear distribution or not). However, this assumes you have a significant training set.

You could also try with a simple genetic algorithm to obtain an mathematical expression that tries to predict the next value based on the previous n. However, this has similar issues to the perceptron approach, as you may lack an adequate size of the training set.

Otherwise, it is probably impossible to predict the future elements without knowing more about the problem. If you know more about the problem, then you can use that information to model the predictor.

Predict future values of time-based dataset in PHP

1 Answers1