Published December 20, 2021
| Version 1.0
Working paper
Open
To what extent can predict Customer Sentiment?
Description
Research Question: Can we create and train a neural network model using NLP techniques to analyze the history of customer reviews and scores to forecast and predict whether customer sentiment is positive or negative?
This study prepares the history of customer reviews and scores and builds and trains the neural network model to predict and forecast customer sentiment and define positive or negative to understand the customer and improve the business by supporting the marketing process.
The study will use the TensorFlow Neural Networks, sentiment analysis is one of the tasks in Natural Language Processing (NLP) techniques that will perform a text classification that can be trained to predict text sequences.
Notes
The study creates and trains a neural network to understand the customer review text to predict Sentiment, classifying positive or negative sentiment based on the words coming in the customer review text. Built the neural network model by sequential of five layers including two hidden layers compile with "Binary Crossentropy" loss function optimized by "Adam" evaluated by "accuracy" metrics and fitted by separated training data and validated by testing data.
Files
D213_Task2_Fady_Missiha.ipynb
Additional details
Related works
- Cites
- Report: https://github.com/fadymissiha/NLP-sentiment/ (URL)
References
- DataCamp (2022, Jan. 19) Sentiment Analysis Nuts and Bolts.
- keras. (2022, Jan. 19) Text classification from scratch.
- Towards. (2022, Jan. 19) Various Optimization Algorithms For Training Neural Network.
- kavita. (2022, Jan. 18) How to Use Tfidftransformer & Tfidfvectorizer.