Meet Bishnusaran Sahu! A Project Manager with 14 years of IT experience in Service Delivery, product consultation, and analytics. Read to learn more about Bishnusaran Sahu’s journey with Great Learning’s PGP Artificial Intelligence and Machine Learning Course in his own words!
During these days, on many occasions in multiple engagements, we see the following common scenario, where the user reports a problem or incident to the Help Desk or Service Desk to get further assistance. Help Desk team categorizes those issues based on the symptom and designates associated remediation based on this. The integral point here is to set the proper priority to the tickets for better utilization of support executives. In many cases, it’s found that the priority of the tickets is influenced by the user, the way it’s explained the urgency and interpretation of help desk engineers. There is a huge scope of misjudgment of the ticket priority. What if we can predict the priority of an issue where the system recommends the right priority and associated remediation plan? This is the problem statement we are trying to address, as explained below.
Using Natural Language Processing (NLP) technique, we can evaluate the historical work notes recorded while solving an incident or problem and associated priority (i.e., P1/P2/P3/P4 or P5) of the tickets. Building a pattern by assessing the historical work notes along with a short or brief description of the tickets allows classifying the tickets with proper priority.
First, understand the dataset. After cleaning up, the data evaluate the feature dependency on the target priority classes and eliminate the unwanted features from the dataset. Apply stemming, lemmatization, remove stop words from the ticket work notes (as needed) and then split the dataset into train and test sets. Apply regular NLP processing techniques, LSTM, Bi-LSTM models, other supervised learning techniques, e.g., SVM, KNN, etc., to evaluate the model performance. Tune the hyperparameter to enhance the model performance and identify the best performing model for classification.
Once the model is deployed into production through a Web Application or under a ChatBot where the user just describes a problem or incident, the system just predicts the recommended priority of the ticket. The business benefit is realized when support staff chooses the right steps to remediate the issue with proper urgency.
