Let’s say you have the following 5 numbers:
values = [100, 200, 300, 400, 500]
To predict the next 3 in this series you need linear regression:
def linear_regression(pairs)
n = pairs.size
sum_x = pairs.map { |x, _| x }.sum.to_f
sum_y = pairs.map { |_, y| y }.sum.to_f
sum_xx = pairs.map { |x, _| x * x }.sum.to_f
sum_xy = pairs.map { |x, y| x * y }.sum.to_f
a = ( n * sum_xy - sum_x * sum_y ) / (n * sum_xx - sum_x**2)
b = (sum_y * sum_xx - sum_x * sum_xy) / (n * sum_xx - sum_x**2)
[a, b]
end
Call the function above with pairs of [index, value]:
def predict_next(count, values)
pairs = []
values.each_with_index { |e, i| pairs << [i, e] }
a, b = linear_regression(pairs)
n = values.size
result = []
count.times { |i| result << a * (i + n) + b }
result
end
Now you can call the #predict_next/2 function:
values = [100, 200, 300, 400, 500]
pp predict_next(3, values)
It will display your result:
[600.0, 700.0, 800.0]