sonwanesuresh95 / rfm Goto Github PK

View Code? Open in Web Editor NEW

10.0 2.0 1.0 3.52 MB

Python Package for RFM Analysis and Customer Segmentation

License: MIT License

Python 100.00%

rfm-analysis python customer-segmentation

rfm's Introduction

rfm

rfm: Python Package for RFM Analysis and Customer Segmentation

Info

rfm is a Python package that provides recency, frequency, monetary analysis results for a certain transactional dataset within a snap. Its flexible structure and multiple automated functionalities provide easy and intuitive approach to RFM Analysis in an automated fashion. It aims to be a ready-made python package with high-level and quick prototyping. On practical hand, real world data is easily suited and adapted by the package. Additionally, it can make colorful, intuitive graphs using a matplotlib backend without breaking a sweat.

Installation

Dependencies

Python (>=3.7)
Pandas (>=1.2.4)
NumPy (>=1.20.1)
matplotlib (>=3.3.4)

To install the current release (Ubuntu and Windows):

$ pip install rfm

Usage

# predefine a transaction dataset as df

>>> from rfm import RFM

>>> r = RFM(df, customer_id='CustomerID', transaction_date='InvoiceDate', amount='Amount')

>>> r.plot_segment_distribution()

License

MIT

Documentation

<-- Temporarily Hosted Here -->

Initialization

Read required dataframe

>>> df = pd.read_csv('~./data.csv')

Import RFM package and start rfm analysis automatically:

>>> from rfm import RFM

>>> r = RFM(df, customer_id='CustomerID', transaction_date='InvoiceDate', amount='Amount') 

>>> r.rfm_table

If you want to do rfm analysis manually:

>>> r = RFM(df, customer_id='CustomerID', transaction_date='InvoiceDate', amount='Amount', automated=False)

Attributes

RFM.rfm_table

returns resultant rfm table df generated with recency, frequency & monetary values and scores along with segments

>>> r.rfm_table

RFM.segment_table

returns segment table df with 10 unique categories i.e. Champions, Loyal Accounts etc.

>>> r.segment_table

Methods

RFM.plot_rfm_histograms()

Plots recency, frequency and monetary histograms in a single row

>>> r.plot_rfm_histograms()

RFM.plot_rfm_order_distribution()

Plots orders by customer number

>>> r.plot_rfm_order_distribution()

RFM.plot_versace_plot(column1, column2)

Plots scatterplot of two input columns

>>> r.plot_versace_plot(column1='recency',column2='monetary_value')

>>> r.plot_versace_plot(column1='recency',column2='frequency')

>>> r.plot_versace_plot(column1='frequency',column2='monetary_value')

RFM.plot_distribution_by_segment(column, take)

Plots Distribution of input column by segment

>>> r.plot_distribution_by_segment(column='recency',take='median')

>>> r.plot_distribution_by_segment(column='frequency',take='median')

>>> r.plot_distribution_by_segment(column='monetary_value',take='median')

RFM.plot_column_distribution(column)

Plots column distribution of input column

>>> r.plot_column_distribution(column='recency')

>>> r.plot_column_distribution(column='frequency')

>>> r.plot_column_distribution(column='monetary_value')

RFM.plot_segment_distribution()

>>> r.plot_segment_distribution()

Plots Segment Distribution, i.e. Segments vs no. of customers

RFM.find_customers(segment)

returns rfm results df with input category

>>> r.find_customers('Champions')

rfm's People

Contributors

Stargazers

Watchers

Forkers

fyx2019

rfm's Issues

Clarifying how the input data should look like

Hi, LOVE this library. Super helpful.

Only problem I had was with the date format :(. It took me a while to decipher how the input format for date ought to look like and then to transform the data.

The solution that I had for my input data which looked like this: (str) 2022-06-22 was
df['date']= pd.to_datetime(df['date'])
df['new_formatted_date'] = df.date.dt.strftime('%m/%d/%y %H:%M')

How to solve the issue?
I recommend to either:
A) Clarify exactly how the input data needs to look like
B) Enable multiple common date type input and transform them on the back-end

Need to create into 4 bins instead of 5

Hi this rfm is constructed for 5bins. I need to do it for 4 bins. How I can proceed

TypeError: unsupported operand type(s) for -: 'str' and 'str' when subtracting `datetime64[ns]`

During RFM calculation, datetime64[ns] strangely becomes str. MWE:

from rfm import RFM
import pandas as pd

df = pd.DataFrame({'timestamp': pd.to_datetime(['01-06-2018 23:15:00',  # Creating data
                                                '02-09-2019 01:48:00',
                                                '08-06-2020 13:20:00',
                                                '07-03-2021 14:50:00']),
                    'amount': range(0, 4),
                    'customerid': range(0, 4)})
print(df.dtypes)
print(type(df['timestamp'].max()))
r = RFM(df, customer_id='customerid', transaction_date='timestamp', amount='amount')
print(p.segment_table)

outputs

timestamp     datetime64[ns]
amount                 int64
customerid             int64
dtype: object
<class 'pandas._libs.tslibs.timestamps.Timestamp'>

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
Cell In [2], line 11
      4 df = pd.DataFrame({'timestamp': pd.to_datetime(['01-06-2018 23:15:00',  # Creating data
      5                                                 '02-09-2019 01:48:00',
      6                                                 '08-06-2020 13:20:00',
      7                                                 '07-03-2021 14:50:00']),
      8                     'amount': range(0, 4),
      9                     'customerid': range(0, 4)})
     10 print(df.dtypes)
---> 11 r = RFM(df, customer_id='customerid', transaction_date='timestamp', amount='amount')
     12 print(p.segment_table)

File ~/Library/Python/3.10/lib/python/site-packages/rfm/rfm.py:30, in RFM.__init__(self, df, customer_id, transaction_date, amount, automated)
     28 # automated operations
     29 if automated:
---> 30     df_grp = self.produce_rfm_dateset(self.df)
     31     df_grp = self.calculate_rfm_score(df_grp)
     32     self.rfm_table = self.find_segments(df_grp)

File ~/Library/Python/3.10/lib/python/site-packages/rfm/rfm.py:70, in RFM.produce_rfm_dateset(self, df)
     67 # finding r,f,m values
     69 latest_date = df[self.transaction_date].max()
---> 70 df_grp['recency'] = df_grp[self.transaction_date].apply(lambda x: (latest_date - x[-1]).days)
     71 df_grp['frequency'] = df_grp[self.amount].apply(len)
     72 df_grp['monetary_value'] = df_grp[self.amount].apply(sum)

File ~/Library/Python/3.10/lib/python/site-packages/pandas/core/series.py:4771, in Series.apply(self, func, convert_dtype, args, **kwargs)
   4661 def apply(
   4662     self,
   4663     func: AggFuncType,
   (...)
   4666     **kwargs,
   4667 ) -> DataFrame | Series:
   4668     """
   4669     Invoke function on values of Series.
   4670 
   (...)
   4769     dtype: float64
   4770     """
-> 4771     return SeriesApply(self, func, convert_dtype, args, kwargs).apply()

File ~/Library/Python/3.10/lib/python/site-packages/pandas/core/apply.py:1105, in SeriesApply.apply(self)
   1102     return self.apply_str()
   1104 # self.f is Callable
-> 1105 return self.apply_standard()

File ~/Library/Python/3.10/lib/python/site-packages/pandas/core/apply.py:1156, in SeriesApply.apply_standard(self)
   1154     else:
   1155         values = obj.astype(object)._values
-> 1156         mapped = lib.map_infer(
   1157             values,
   1158             f,
   1159             convert=self.convert_dtype,
   1160         )
   1162 if len(mapped) and isinstance(mapped[0], ABCSeries):
   1163     # GH#43986 Need to do list(mapped) in order to get treated as nested
   1164     #  See also GH#25959 regarding EA support
   1165     return obj._constructor_expanddim(list(mapped), index=obj.index)

File ~/Library/Python/3.10/lib/python/site-packages/pandas/_libs/lib.pyx:2918, in pandas._libs.lib.map_infer()

File ~/Library/Python/3.10/lib/python/site-packages/rfm/rfm.py:70, in RFM.produce_rfm_dateset.<locals>.<lambda>(x)
     67 # finding r,f,m values
     69 latest_date = df[self.transaction_date].max()
---> 70 df_grp['recency'] = df_grp[self.transaction_date].apply(lambda x: (latest_date - x[-1]).days)
     71 df_grp['frequency'] = df_grp[self.amount].apply(len)
     72 df_grp['monetary_value'] = df_grp[self.amount].apply(sum)

TypeError: unsupported operand type(s) for -: 'str' and 'str'

documentation for useful function

I see that there are lot of attributes to r object. Meaning if I type r. it populates lot of methods like dynamic cut offs, dynamic rfm score etc. It would be really useful to have documentation for these methods. So, we can experiment/customize thresholds based on our requirements

rfm object changes `NaN` to `nan`

I have a column in my dataframe (that is not used in rfm analysis) like as below

However, once I execute the below code

r = RFM(df_new, customer_id='unique_key', transaction_date='Date', amount='Revenue')

the same column market segment MC becomes like as below. I verified multiple times. Not sure why is this happening. Is there any shallow copy, deep copy etc stuff happening that is causing change in my input dataframe?

segment criteria - RFM

Hi,

I recently came across this package and it is very useful package. Helps people who lack extensive coding knowledge. Appreciate it.

However, I have one quick question.

How can I find out the definition and RFM score criteria for each segment?

For ex: I get a graph like below and how do I customers with which score go into Champions segment?

Can direct me to a documentation where I can find the definition and score criteria for each segment? There is no documentation link in your readme. page

Meaning, if it is 555 it is champions. What if it is 541? is it still considered champions? So, I would like to know the cut off points that you chose for each segment.

`How much drop in revenue value or recency value or frequency value is required for a record to be put into another segment? Is it 100$ or 2 days for recency or 2 points for frequency?

Basically, am looking for a table like below which I got from tutorial https://www.analyticseducator.com/Blog/Customer_Segmentation_Using_RFM_with_Python.html

Additionally, is this a new package? Can it be used for real world business insights or is it still under testing?

Nonetheless, great effort and useful package for people like me

Customers in multiple segments - expected?

I was going through your code and pasting a small section of it

 if (r in (4,5)) & (f in (4,5)) & (m in (4,5)):
                classes_append({rec[cust]:'Champions'})
            elif (r in (4,5)) & (f in (1,2)) & (m in (3,4,5)):
                classes_append({rec[cust]:'Promising'})
            elif (r in (3,4,5)) & (f in (3,4,5)) & (m in (3,4,5)):
                classes_append({rec[cust]:'Loyal Accounts'})

Based on this definition, a customer who is a Champion will also be put under the Loyal accounts category. Am I right? Is this how RFM works as well? I thought each customer will only be in one segment only

So, it is possible that same customer will be present in multiple segment?