Python - how to determine the "trend" of data?

Question

Python

Good evening . There was a need to find a trend (growth / decline) of data in Python. The trend must be found almost in real-time (maximum delay of 1 s).
The variable float is given at the input every 0.05 s, at the output I want to get the directions of the graph (trend) - growing (a noticeable clear increase in peaks), decreasing (a clear decline in peaks) or without noticeable fluctuations (the peaks differ very little from each other (2 - 5%) , "stable") . I tried to apply the method of least squares () and other variants of regression analysis, but due to poor skills in Python, I could not implement it.
I would be grateful for any help on this subject.

Answer 1

Look here
stackoverflow.com/questions/7171356/python-calcula...

Answer 2

What went wrong with the least squares method?
Here is the simplest linear regression solution:

def linregress(x, y, w=None, b=None):
    x = np.array(x, dtype=np.float64)
    y = np.array(y, dtype=np.float64)

    if w is None:
        w = np.ones(x.size, dtype=np.float64)

    wxy = np.sum(w*y*x)
    wx = np.sum(w*x)
    wy = np.sum(w*y)
    wx2 = np.sum(w*x*x)
    sw = np.sum(w)

    den = wx2*sw - wx*wx

    if den == 0:
        den = np.finfo(np.float64).eps

    if b is None:
        k = (sw*wxy - wx*wy) / den
        b = (wy - k*wx) / sw
    else:
        k = (wxy - wx*b) / wx2

    return k, b

The data vectors x and y are input (you can also set the weight vector), the output is the coefficients k, b for the model y = kx + b.
"Magic" functions in numpy: polyfit/polyval.
Error calculation functions, such as sse/mse/rmse/rsquare, are easy to write using formulas that can be found on the Internet or in books on statistical and regression analysis.