Introduction About Data

This is the random data of the sales by customeracross all differnet categories of the products offered at Firm ABC Corp.

Data Fileds

Field NAME Description Note
Customer Customer More than one record in the data More than One record for each Customer
Brand Category Name of the Products More than One record for each Customer
Order_Qty Order Quantity
Orer_Cost $ Value of Order
city City of the Customer
state State of the Customer Missing for few Entries
Year Year of the Order
Country Country

Sample

Customer Brand Order_Qty Orer_Cost city state Year Country
1221893 RB 19 1663.55 CENTREVILLE AL 2018 US
1221893 RX 22 1486.71 CENTREVILLE AL 2018 US
1221893 RX 8 584.28 CENTREVILLE AL 2018 US
1221893 RY 4 158.68 CENTREVILLE AL 2018 US
1221894 RB 2 157.46 ALABASTER AL 2018 US
1221894 RX 11 794.18 ALABASTER AL 2018 US
1221894 RY 5 200.27 ALABASTER AL 2018 US
1221894 RX 11 733.56 ALABASTER AL 2018 US
1221894 RY 3 120.89 ALABASTER AL 2018 US

Analytical Framework

Domain

Opening a New Category for the Customer.

Purpose

  1. Identify Best possible Brand Customer can be proposed given they are selling Brand A,B, and C or More.
  2. We need to consider State as 2nd variable which could impact the proposal as one of the attributes.
    For example, I might be able to propose customer selling A from NY to start selling Brand BB, but not to as Customer Y from CL, as there is no enough data to prove the sales of this brand at the given state level.
  3. Recommendation would take Brand already carried by Customer to Propose the result with top 5 Brands.
  4. Building a shiny app to enter the Brand and showing Result in User friendly way by applying different algorithms.

Recommendation Context

When Customers are looking to Expand to new Category or marketing Events where customers can be targeted with new set of Category they should start selling.

Personalization Level

Based on Brand they sales today which other Brand can help them increase the revenue .

Interface

Marketing Events, Agents reaching out to customers and trying to convince them to start the selling the new category with data we already have.

Whose Opinions

We are relaying on data from the Other customers and sales made by them in correlation with other category.

Privacy and Trustworthiness

This data is not real, and objective is to build the system and churn it to real business use.

Recommendation Algorithm [Discuss within team]

  1. SVD / FSVD
  2. USER to USER
  3. ITEM to ITEM
  4. .
  5. .
  6. .

Build, Train and Test Data

Variable Name Desc.
df_ <> For Dataframe
mod_<> For Model Name
df_test_<> For Test Data
df_train <> For Train Data
df_full <> For full data
df_sample_n<> For Sample of Size n
. .
. .
. .

Algorithm A

Algorithm B

Algorithm C

Evaluation of Algorithm

Shiny App

Conclusion

Reference