README

This is a Python implementation based on the ChannelAttribution package in R developed by Davide Altomare and David Loris.

https://cran.r-project.org/web/packages/ChannelAttribution/ChannelAttribution.pdf

Please Note: The authors of the original R library now also provide a python implementation of ChannelAttribution which you should probably use instead of my implmentation since theirs has additional features not available in this package.

Installation

pip install pychattr

Markov Model

import pandas as pd
from pychattr.channel_attribution import MarkovModel

data = {
    "path": [
        "A >>> B >>> A >>> B >>> B >>> A",
        "A >>> B >>> B >>> A >>> A",
        "A >>> A"
    ],
    "conversions": [1, 1, 1],
    "revenue": [1, 1, 1],
    "cost": [1, 1, 1]
}

df = pd.DataFrame(data)

path_feature="path"
conversion_feature="conversions"
null_feature=None
revenue_feature="revenue"
cost_feature="cost"
separator=">>>"
k_order=1
n_simulations=10000
max_steps=None
return_transition_probs=True
random_state=26

# instantiate the model
mm = MarkovModel(path_feature=path_feature,
                 conversion_feature=conversion_feature,
                 null_feature=null_feature,
                 revenue_feature=revenue_feature,
                 cost_feature=cost_feature,
                 separator=separator,
                 k_order=k_order,
                 n_simulations=n_simulations,
                 max_steps=max_steps,
                 return_transition_probs=return_transition_probs,
                 random_state=random_state)

# fit the model
mm.fit(df)

# view the simulation results
print(mm.attribution_model_)

  channel_name  total_conversions
0            A           1.991106
1            B           1.008894

# view the transition matrix
print(mm.transition_matrix_)

  channel_from    channel_to  transition_probability
0      (start)             A                     1.0
1            A             B                     0.5
2            A  (conversion)                     0.5
3            B             A                     1.0

# view the removal effects
print(mm.removal_effects_)

  channel_name  removal_effect
0            A          1.0000
1            B          0.5067

Heuristic Model

import pandas as pd
from pychattr.channel_attribution import HeuristicModel

data = {
    "path": [
        "A >>> B >>> A >>> B >>> B >>> A",
        "A >>> B >>> B >>> A >>> A",
        "A >>> A"
    ],
    "conversions": [1, 1, 1],
    "revenue": [1, 1, 1],
    "cost": [1, 1, 1]
}

df = pd.DataFrame(data)

path_feature="path"
conversion_feature="conversions"
null_feature=None
revenue_feature="revenue"
cost_feature="cost"
separator=">>>"
first_touch=True
last_touch=True
linear_touch=True
ensemble_results=True

# instantiate the model
hm = HeuristicModel(path_feature=path_feature,
                    conversion_feature=conversion_feature,
                    null_feature=null_feature,
                    revenue_feature=revenue_feature,
                    cost_feature=cost_feature,
                    separator=separator,
                    first_touch=first_touch,
                    last_touch=last_touch,
                    linear_touch=linear_touch,
                    ensemble_results=ensemble_results)

# fit the model
hm.fit(df)

# view the heuristic results
print(hm.attribution_model_)

  channel  first_touch_conversions  ...  ensemble_revenue  ensemble_cost
0       A                      3.0  ...               8.1            5.1
1       B                      0.0  ...               0.9            0.9
[2 rows x 13 columns]