Holt Forecast

Problem

Fit Holt's linear trend model with double exponential smoothing and forecast h steps ahead as level + h * trend. Seed the states with level = x[0] and trend = x[1] - x[0], and use the usual updates new_level = alpha * x[t] + (1 - alpha) * (level + trend) and new_trend = beta * (new_level - level) + (1 - beta) * trend.

A single-element series returns x[0]; an empty series returns 0.0.

Hint

Extrapolate the trend linearly: each extra step adds one more b.

Test cases

case 1

in: [[1,2,3,4,5],0.5,0.5,1]

expected: 6

case 2

in: [[1,2,3,4,5],0.5,0.5,2]

expected: 7

case 3

in: [[10,12,14,16],0.3,0.2,1]

expected: 18

case 4

in: [[10,12,14,16],0.5,0.5,3]

expected: 22

Starter codePython

def holt_forecast(series, alpha, beta, h):
    # Your code here
    pass

4 test cases · run your code against them in the editor