Is there any way to interpolate in 2D with pandas?

Question

My dataframe looks like:

X   Y   Val
10  10  17
10  30  28
10  40  35
10  50  15
20  10  17
20  30  18
20  40  33
20  50  15
40  10  37
40  30  29
40  40  49
40  50  40

I want to extract a interpolated "Val" for given "X" and "Y" even if they are not in the df.
For example: I can very easily get "Val" for X=10 and Y=50.

But, suppose I want X=15 and Y=15 which is not in dataframe? Is it possible to extract the same? Does Pandas have any such functionality?

Answer 1

I don't know about pandas, but you can do it through scipy.interpolate.interp2d:

f = interp2d(x = df['X'], y = df['Y'], z = df['Val'])

Then you can check the value of the function in a given point (x, y):

>>> f(15, 15)
array([17.97919182])

Answer 2

As an entry point to machine learning, this could be a good way to experiment with a linear regression model:

import pandas as pd
from sklearn.linear_model import LinearRegression

# create a sample dataframe
data = {'X': [10, 20, 30, 40], 'Y': [10, 20, 30, 40], 'Val': [100, 200, 300, 400]}
df = pd.DataFrame.from_dict(data)

# define the target and input data for training
target = df.Val
df.drop(['Val'], axis=1, inplace=True)
input_df = df

# fit the linear regression model
regression_model = LinearRegression()
regression_model.fit(input, target)

# create another sample dataframe to test output
prediction_data = {'X': [50, 60, 70], 'Y': [50, 60, 70]}
prediction_df = pd.DataFrame.from_dict(prediction_data)

# make predictions
regression_model.predict(prediction_df)

Which gives the output:

array([500., 600., 700.])