Customer Connectivity Issues needs a pro-active solution. For our approach on building an ML model, the first step is to utilize GCP data prep for cleansing, feature engineering, etc.
Business Problem – Connectivity Issue
- The customer has multiple Enterprise to Store connections for the exchange of information but is very critical for the Customer’s business. Currently, there are connectivity issues between Enterprise(EAI Layer) and the Stores(Store Layer) across every region
- Majority of the time the error that we are getting is a connection timeout because at the Stores the Firewall rules were not configured properly or either they missed adding firewall rules as per the network playbook to accept the traffic to/from the Stores and Enterprise
- Script to automate connectivity check has been implemented for a month which provides a log file with timestamp and results of the successful or unsuccessful transaction
- With the number of stores that Customer has, it is becoming a tedious and challenging task for our team to ensure the connectivity to/from the stores. Pro-active measures would be very helpful
The connectivity results are generate in the Log Files that will be used as our input dataset for ML model to predict the issues
- Step 1 – Use Google’s Dataprep to prepare the data for the ML Model (Cleansing, Feature Engineering, etc.)
- Step 2 – Use multiple approach and algorithms to find the best suited ML model for the use case
- Step 3 – Deploy the algorithm and predict the failures that the project team can use for pro-actively monitoring the connections
As we know, 70-80% effort for the ML model is done for data prepping. Our dataset also needs cleansing, reformatting, and other conditional formatting, which is done in a few easy steps by using dataprep.
Once the data prep recipe is created, it can be scheduled and/or reused in the future.