List of possibles thoughts a customer can have
Inspiration Me being a customer I want the bank to know me well. I would like to have a very own version of bank of west website that pulls all relevant data that I would like to see or that Im searching for. Advertise me with the stuff that Im most interested in, remind me often on the product updates that I need . Bank being able to restructure its entire website with machine learning models for me.
What it does
We went through bank of west website. We found a a lot of interconnected things (webpages).
We decided to build an AI model that can understand the customer, what he thinks what he wants , connects all of the webpages of bank of west.
By knowing the customer The model will be able to restructure the entire webpage of bank of west specific to that customer for rich customer experience Pair products with people for advertisement
With past history of data say from JAN - DEC (2018) of me, the model will be able to predict what I would need on JAN 1 2019
How we built it
KEY TERMS : Python, SCI - KIT (MUTLI NOMIAL NAIVE BAYES with Count vectorisation approach), Bi-directional search, Linked Data, Sentimental analysis. We pulled all the training data sets from bank of west website, we linked the training dataset into tree structure.
To build the model:
We cleaned the data set removed stop words, words that had very less correlation with the entire dataset (i.e the words that affected the accuracy). We used count vectorisation approach to model the entire dataset into two dimensional matrix. We applied multinomial naive bayes with default alpha (learning rate) and predicted the class with the highest probability, correlation to the search query of the user
We built a user interface that receives thoughts from the user and send it to the server to fetch all the relevant data With the response from the server we dynamically constructed the webpage that a user would want to see with all the relevant data that he is looking for or wants to see using java script, html css. We built a user interface for bankers to upload a new product idea and find all the potential customers for that product
Front End to Back End : We ran a server in port 5000 in python and connected our java script as a client to it so that we can transfer data across .
Apart from the model, we built bidirectional search that would help the model to collect data related to what the user is looking for and provide additional information that would help the customer. We also did sentimental analysis on the comments to get the top rated positive comments related to the user search
Challenges we ran into
We faced challenges on improving the accuracy of the model from being smart to perfection on every search result. Come up with good results for some of the most difficult un meaningful queries.
Accomplishments that we're proud of
We not only made a complete working model but we were able to come up with a model with good accuracy. Built UI specific to the customer
What we learned
A lot of banking approach. A different problem statement (know you customer) and its solution using a machine learning approach.
What's next for (KYC) Know Your Customer
We have lots of ideas to improve the idea. Build a strong core module which is open for extension