Abstract:
This thesis introduces an advanced Natural Language Processing (NLP) framework for in-depth
analysis of customer behavior, specifically focusing on Customer Comments of social media in
Bengali (CCSMB). The study leverages a variety of cutting-edge algorithms, including Recurrent
Neural Networks (RNN), Long Short-Term Memory (LSTM), k-Nearest Neighbors (KNN),
Logistic Regression, and Decision Trees, combined with rigorous data cleaning techniques tailored
to the linguistic intricacies of Bengali text.
The study utilizes a meticulously curated dataset from diverse social media platforms, ensuring the
robustness and adaptability of the models to the dynamic nature of online interactions. Evaluation
metrics such as accuracy, precision, recall, and F1 score are employed to assess the performance
of each algorithm in capturing text emotion detection and categorizing topics within the Bengali
linguistic context.
The research begins with an extensive exploration of data cleaning methodologies, addressing
challenges such as noise, irrelevant information, and linguistic nuances unique to Bengali.
Subsequently, various machine learning and deep learning algorithms are applied to the
preprocessed data. RNN and LSTM models are utilized for sequential analysis of customer
comments, capturing temporal dependencies in the expression of sentiments.
Analysis and result
The results not only contribute to the evolving landscape of customer behavior analysis but also
offer practical insights for businesses seeking to enhance customer engagement and satisfaction in
the Bengali-speaking market.
This research underscores the versatility of NLP techniques and a multitude of algorithms in
unraveling valuable insights from customer interactions, emphasizing their applicability in
fostering customer-centric strategies in an increasingly digital and multilingual business
environment.
