Proposed Idea Template Created by Dr
Proposed Idea Template
Dr. Ammar El-Adl
Student Name Course Name Initial Date Version Version Date
AbdelsadekGraduation project 1.0 1.0 7/24/2018
Item Description & Steps
Title Tracking how a change in a service affects Telecom Customers feeling using Sentiment Analysis ‘Naïve Bayes’.
Abstract Context of the problem Telecom companies get a bad rap when it comes to customer experience. All too often, clients feel that service falls short of their expectations, and that complaints seem to be falling on deaf ears. Yet despite poor customer sentiment, few telecom companies have made customer centricity a priority, so we try to use Machine Learning Sentiment Analysis to determine if the Telecom Customer is happy or not with the new change in a service; Then we can produce the best services to the customers. The sentiment Analysis is the process of computationally identifying and categorizing opinions expressed in a piece of text, especially in order to determine whether the writer’s attitude towards a particular topic, product, etc. is positive, negative, or neutral. As the business changes so does the customer sentiment. Publishing a marketing campaign or press release, changing your service’s interface or price structure can have an effect. Tracking customer sentiment can measure this.
Task Find a conversation chat bot Dataset, Implement the Naïve Bayes classifier, Split the data into train set and test set, Train the classifier with training set, Test it with testing set, give the customer a new service, Get the text of the conversation from chat bot, Compute the sentiment of the customer, Try to help him if he/she was not happy with the change in the service.
Try to improve the new service if the customers were willing and finding a way to update or change it if they were not happy.
Objective Determine if the customers are willing with the new service or not, For helping the Telecom customers to find the best services, Increasing the Telecom customers, improving of the Telecom Customer care and helping Telecom company to determine the best services to improve it.
Importance (Why?) By analyzing the sentiment more accurately, and in particular finding the things people are really unhappy about, you can :
Focus more on what will make a difference
help users to find the need and best service
Help the company to produce the best service for its customers.
Help company for improve its customer care by keep track what the percentage of acceptance that the new service generates.
If you know aspects and themes in each response, you can also answer questions like: For how long do people react negatively to a change in a service, or do they really love the new feature added?
Uniqueness Proof Previous work There are no Previous works on the same direction. But there are some works that similar to this work like Machine Learning-Based Sentiment Analysis for Twitter Accounts1 , Sentiment Analysis of Review Datasets using Naïve Bayes’ and K-NN Classifier2, understanding customer experience3
The problem Name How to Know if the Telecom Customers are willing or not with the new service? If not, How to help them?
Premises We will collect a conversations chat bot dataset for training and testing the Naïve Bayes Classifier. Change a service and connect to the customers using the Chat Bot, Tack the customer conversation to calculate its sentiment (positive or negative) After taking the body of the conversation into some of the Context Analysis and Text Preparation for removing stop words ,Convert the slang words to its general meaning ,Computing the Bag of words and so on. Then we can determine if the new service is helpful or not.
Tools We may not use any tools.
Values Description: The values are to be used to solve this problem.
List the values used to solve the problem.
Methods To solve this problem we need to some Methods like the Machine Learning Sentiment Analysis “Naïve Bayes”, Natural Language Processing “Bag of Words, Tf-Idf” and Some of Context Analysis Techniques
Core Determine if the Telecom customers are willing or not with the new service.
Questions There are a set of Questions that need to be answered to solve the problem such as: Does the Telecom Company needs to offer a new service? If The Telecom Company change one of The Current service or adding a new Service how long people react negatively to the new service do they really love the new feature added? How can we help the company for offering the best services? How can we help the Telecom company to reach to the customer needs?
Proposed solution Name Using ML Sentiment Analysis ‘Naïve Bayes’ to determine the success of the new services.
Assumptions We will find the conversation dataset then implement the naïve Bayes to classify if the text is Negative or Positive. We may need to use IBM Watson Personality Insights for computing the big five traits based on the passed conversation text, this will help us for deeply identify if the Telecom customers are willing or not.
We will use the Telecom chat bot for communication with the Telecom customers, Then we may try to offer him a new service then we tack the customer reply for identifying if he was willing or not , then we give the company part a report about the acceptance of the new service ,by this way the company can easily help its customers by providing the best service for its customers.
Obstacles ; Constraints We may have less data because the Telecom company can’t give you its customer data. We may need machine that would be faster because of the Deep learning need to very fast devices. We may have a problem with the slang speech as not all people speech with the general formal language but they usually use the slang.
Solution path Name Using Sentiment Analysis “Naïve Bays” and NLP “Bag of words, Context Analysis and TF-IDF” for offering the willing Telecom services.
Configurations Description: Constants and constraints values used or assumed to run this path.
Use the assumptions to extract configurations values.
Explain why those values were used.
Stages We need to collect the Telecom Conversation customers that have a Negative and Positive conversations.
Move the data set through text analysis for removing stop words.
Computing the bag of words.
Then we need to divide the dataset into the Training and Testing sets based the k-folds algorithm, this steep will return the division with the best manner(percentage)
Then we need to implement the Naïve Bayes Algorithm and training and testing it with the training and testing sets.
We need to change any service or offering a new service.
Communicating with the customer through chat bot.
Then we need to notes the customer reaction by tacking the string of their conversations to classify if they are happy or not.
We should give the Telecom a full report about the customer willing.
Verification Separate the data into training and testing sets to test the accuracy of predictions. And Ask business experts to review the results of the model to determine whether the discovered patterns have meaning in the targeted business scenario.
Validation We will use the validation set of the dataset for validating the output and insuring that the algorithm give the high accuracy.
Targeted outputs There is no doubt that all of the Telecom Company need to care about its Customers, We try to help it by make a full observation to the customer felling and interaction after each change in a current service or offering a new service, but this only doesn’t useful so We need to try to identify the best and bad services, By this way the company could easy Care about its Customers.
Pros There is no doubt that the “Naïve Bayes Classifier” is one of the best binary classifiers because of using the Probability Statistics for classification, If the NB conditional independence assumption holds, then it will converge quicker than discriminative models like logistic regression. Even if the NB assumption doesn’t hold, it works great in practice. Need less training data. Highly scalable. It scales linearly with the number of predictors and data points. Can be used for both binary and multi-class classification problems.
Can make probabilistic predictions and Handles continuous and discrete data.
By removing the stop words we can reach to the cleaning text that each word on it has a scientific or emotional meaning that can effect on the classification result. We need to calculate the bag of words and the tf-Idf that help us on knowing the core and the important words on the Text. Using tf-idf Because we focus on counting of words in documents.
Cons Some of the disadvantages of using the “Naïve Bayes” that it depends on a huge dataset and complex computation. Another problem happens due to data scarcity. For any possible value of a feature, you need to estimate a likelihood value by a frequents approach. Bag of words representation doesn’t consider the semantic relation between words, it just focus on count of word and neglect the arrangement, n-grams and tagging in sentence
Path recommendation Solution Recommendation Using the ML SVM may help to reaching to best accuracy because it is one of the best binary classifications.