Abstract:
These days, Artificial Intelligence is playing a key role in the progression of humanity as it helps
to curtail human struggle in every aspect of life. An Immense amount of data is in structured and
non-structured foam from numerous industrial platforms that are striving to get into the shape of
useful information to be a part of scientific research. Although today’s major concern is how to
manage a huge amount of feedback data i.e., Text format citizen complaints. At this point,
proposing a model that automatically classifies the textual complaints by analyzing the content
with the help of NLP (Natural Language Processing) and different ML (machine learning) models
can be beneficial. Primarily, data of complaints are collected from the concerned platforms as well
as from the international Consumer Complaint Database (for validation). The methodology is
comprised of four different stages i.e. (1) initial pre-processing (2) preprocessing (3) future
extraction (a) count vectorizer (b) term frequency-inverse document frequency (TF-IDF) (4) ML
models for categorical classification of the complaints. At the evaluation stage, 10 different classes
are present in assembled complaint dataset and more than 70 % accuracy is achieved from all
classifiers. Likewise, on Consumer Complaint Dataset, 86% accuracy has been achieved. This
model is used to optimize the complaint division automatically and saves a lot of time.
