zolabar/trendpy

Time Series Regression with Python

trendpy's Issues

problems with trigReg function in methods.py

When testing the different regression approaches from methods.py in my local version of the TrendPy Webapp linear and polynomial worked fine. However the output of the trigReg function seems not to work correctly. Coefficients and trendline when using trigReg can be seen in the screenshot below.

Gelbbrustara malen, als Projektlogo

Eigenen Gelbbrustrara als Logo fürs Projekt malen.

Auf Papier und fotografieren oder gleich digital.

Adding Binder link to launch the TrendPy Webapp via Voila

Would be great to have a Binder link to launch the TrendPy Webapp via Voila like you can for example in the conformal maps project, so that everyone can easily use the app.

Web App in Streamlit

Rebuilt the voila app as streamlit app and offer both

include expReg function into methods.py

Exponential regression

f(x) = a exp(bx)

can be reduced to linear regression by taking to log (ln).

ln (f) = b*x+ln (a)

hence, expReg can be easily implemented using np.log and linReg

"freeReg" regression

Trend objects could support a SymPy based "free ansatz" regression, for example

reg = Trend(x_in, y_out, "freeReg","aarctan(bx)")

"aarctan(bx)" could be converted by SymPy to a numerical object and in combination with curve_fit from SciPy the fir could be simply done.

A function in methdods.py called freeReg(x, y, "aarctan(bx)") is the start.

freeReg and initial condition

• give x0 (starting value as option for the free reg part)
• by default try ones and zeros and take the best of both

update documentation

• documentation building connection to code
• Include some theory?

double names

• predict additionally to pred
• coefs additionally to coef