What approach would work well for predicting earthquake intensity based on historical data?

Question

My problem: I own warning system where I collect data from institutions and send them over through various ways to users. I would like to hear your advice on what approach I can use for solving my problem with earthquake intensity far from epicenter. Since seismogical institutions mostly issue info about intenstiy of an earthquake for the epicenter, I would need to predict and classify what intensity the earthquake can have for places distant of several km/miles from the epicenter.

As an input/training set, I can use data of historical earthquakes and their magnitudes in an epicenter. Then I would need to fill mostly "by hand" an information about intensity based on seismological records, historical testimonies, chronicles atc.

What I need from AI: I need "something" that would predict earthquake intensity based on dataset of historical earthquakes.

Example/TLDR: There is an earthquake with magnitude 3.8, distant 80 km with depth 6 km. Based on dataset of historical earthquakes (with same type of information + witnessed and collected intensity), and output, I would need prediction of intensity of an eartquake 80 km from the epicenter.

Answer 1

You can do prediction like this with regression. Take a look at scikit-learn models for regression: https://scikit-learn.org/stable/modules/classes.html#module-sklearn.linear_model

Here's an example:

# This is made up data!
#        epicenter magnitude, depth_m, distance_km, local_magnitude
data = [[3.5,                 6000,    61.3,        2.1],
        [9.1,                 16000,   261.0,       6.0],
        [9.1,                 16000,   96.5,        8.2],
        [3.8,                 6000,    80.0,        2.6],
        [2.5,                 4000,    101.3,       0.4],
        [7.0,                 9000,    81.9,        6.1],
        [5.1,                 7000,    21.3,        4.1]]

from sklearn.linear_model import LinearRegression
import numpy as np

data = np.asarray(data) # to numpy array
X=data[:,:3] # the input variables
y=data[:,3] # the output variable (local magnitude)
reg = LinearRegression().fit(X, y) # fit the model based on the data
reg.predict(np.array([[6.6, 8000, 27]])) # Get a prediction based on new inputs

>>> 6.1511027