Background History
The reason behind multiclass feedback classification models
Historically, the task of categorizing customer feedback in the FastTrack Feedback Management Tool has proven to be a labor-intensive process. It requires manual labeling of feedback into one of four categories, which ultimately diverts resources from two teams who analyze, categorize, and drive necessary action. With multiclass feedback classification models, those resources can be utilized to focus on customers.
Problem Statements and Benefits
Strategic thinking
The existing process of feedback categorization is not only labor-intensive but also time-consuming for our human analysts. By automating this process using multiclass feedback classification models, we anticipate a substantial reduction in manual effort, freeing up valuable human resources for more strategic tasks.
Valuable business outcomes
Furthermore, we are leveraging the manually classified data to train our machine learning models. Once implemented, this solution will significantly expedite the users’ ability to gain insights from the labels and execute actions, yielding valuable business outcomes.
Overcoming challenges
Lastly, the volume of data points (300) in a year along with available data features (3) poses a big challenge for traditional ML algorithms to solve.
Our Solution
Developing multiclass feedback classification models
Our team at GJ has developed classification models using various cutting edge Machine Learning (ML) models including ensemble models, Large Language Models (LLMs), and neural networks. Currently, our models’ accuracies hover around 85%. We are actively working to enhance the value of these evaluation metrics and finalize the model’s deployment with the Dev team.
