Master Thesis
https://onlineresource.ucsy.edu.mm/handle/123456789/2230
2024-03-29T13:44:37ZOPINION MINING SYSTEM OF CUSTOMER REVIEWS BY USING FEATURE EXTRACTION
https://onlineresource.ucsy.edu.mm/handle/123456789/2796
OPINION MINING SYSTEM OF CUSTOMER REVIEWS BY USING FEATURE EXTRACTION
LWIN, NANDAR MOH MOH
Due to the dramatic improvement of ecommerce, web sources which are
important for both potential customers and service providers rapidly emerge in
prediction and decision purposes. Opinion mining techniques become popular to
automatically process customer reviews by extracting features and user opinions
expressed over them. To overcome the task of manual scanning through the large
amount of one-by-one review, people have interested to automatically process the
various reviews and to provide the information which is useful for customers and
service providers. By applying dependency relations, it can properly identify the
semantic relationships between features and opinions of each review. It can find the
numeric score of all the features using SentiWordNet. This system is intended to
collect customer reviews from tourism field and then extract the related features and
opinions to rate the services. Finally, it can rank each agency according to the final
result of each review sentence. In this thesis, Standard Parser is used to generate the
features, opinions and the dependency relations for each trip review at the
preprocessing. The two methods of features extraction such as frequency-based
feature extraction and dependency grammar-based feature extraction are used to
extract the most relevant trip features. Moreover, SentiWordNet 3.0 is used to get the
positive score and negative score for each trip feature and then the system calculates
the total weight of the trip review by using these numeric scores. The objective of the
system is to rank the travel agencies according to the final weight of each travel
agency that is collected by adding the total weight of the trip reviews for that agency.
Therefore, the system implements efficiency and effectiveness in opinion mining to
express the reviewer’s opinion and feeling for next customers’ trip plans by using
features extraction. In this system, Tourism Reviews are applied as the case studies to
identify what elements of an agency affect sales most and what are the features the
customer like or dislike so that trip managers and agency owners can target on those
areas. The system is developed using Java language and MySQL to build the
database.
2023-07-01T00:00:00ZTHE CAR INSURANCE CLAIM PREDICTION SYSTEM BY USING MACHINE LEARNING ALGORITHMS ON APACHE SPARK PLATFORM
https://onlineresource.ucsy.edu.mm/handle/123456789/2795
THE CAR INSURANCE CLAIM PREDICTION SYSTEM BY USING MACHINE LEARNING ALGORITHMS ON APACHE SPARK PLATFORM
Ko, Thein Than
Car insurance companies face a major challenge in dealing with insurance claims,
which are prone to fraud and increasing in volume. This makes it difficult for insurers
to classify claims during the review process. To address this issue, the aim of this study
is to develop four Car Insurance Claim Prediction Classifiers with Random Forest and
Logistic regression based on the car insurance claim dataset respectively and supports
for comparison which method and attributes are more suitable for car insurance
companies. Firstly, this proposed system creates a feature selection model using
Variance Threshold Selector method to select the important attributes impact on the
accuracy of car insurance claim prediction classifiers. The data set is split into training
with 80% and testing sets with 20% randomly and the two classifiers with all attributes,
the training dataset is used to create the LR classifier and RF classifier. For two
classifiers with the feature selection method, the system creates the new training dataset
and new testing dataset by removing low variance value of attributes using Variance
Threshold Selector method. After that, two LR classifier and RF classifier are been
created by using new datasets. The system has analyzed the different attributes: 30, 32,
34, 36, 38, 40 and 42 to choose the number of attributes and important attributes and
tested 10 times for each attribute number because of splitting training and testing
datasets randomly. Finally, the system compares the evaluation results with metrics:
accuracy and f score. RF classifiers with and without the feature selection method are
suitable for the proposed system than LR classifiers. Among different attribute
numbers, the classifiers based on 38 attributes and 40 attributes are the best classifiers
and classifier based on 42 attributes are the second best classifier.
2023-05-01T00:00:00ZCLASSIFICATION OF BANK MARKETING DATA USING SUPPORT VECTOR MACHINE
https://onlineresource.ucsy.edu.mm/handle/123456789/2794
CLASSIFICATION OF BANK MARKETING DATA USING SUPPORT VECTOR MACHINE
Khin, Ei Ei
Nowadays, banking system plays an important role of financial
sectors all over the world. The more accurate predictive modeling system
is required for their services or products in the banking industry. Bank
workers can make those predictive models with manually, but this process
takes long time and lots of man-hours. For these reasons, machine learning
techniques are useful to predict the outcomes with huge amounts of data.
Classification is an important technique to analyze and to predict the data.
This system will implement the classification of bank marketing data using
support vector machine (SVM) to predict the probability of the customers’
subscription to the term deposit whether subscribe or not. Support Vector
Machine (SVM) is a supervised learning model used for classification and
prediction of data. The purpose of this system is to predict the customers'
response to the term 'deposit' using bank marketing data. The precision,
recall, and F-Measure confusion matrix is used to gauge the system's
correctness. In the first experiment when the training data is used, the
accuracy without feature engineering is 86%, the accuracy with feature
engineering is 83% and the accuracy with feature engineering of
Correlation Matrix and Principal Component Analysis gets 96%. In the
second experiment which is used the testing data, the accuracy without
feature engineering gets 85%, the accuracy with feature engineering before
using PCA is 83% and the accuracy after using PCA is 95%. The system
shows the best results in both training data and testing data after using the
Principal Component Analysis.
2023-05-01T00:00:00ZMYANMAR ENTITY IDENTIFICATION FOR NATURAL LANGUAGE UNDERSTANDING USING BIDIRECTIONAL LONG SHORT TERM MEMORY (BiLSTM)
https://onlineresource.ucsy.edu.mm/handle/123456789/2790
MYANMAR ENTITY IDENTIFICATION FOR NATURAL LANGUAGE UNDERSTANDING USING BIDIRECTIONAL LONG SHORT TERM MEMORY (BiLSTM)
PHWAY, SAUNG THAZIN
Entity identification is an exacting function which has commonly
appropriate broad chunk of awareness in the course of feature engineering
and word list to attain great achievement.. Entity Identification (EI) is
indispensable of perceptive article character from basic input and resolve the
division the morphemes characterizes. This paper presents every Entities
Recognition (ER) for Myanmar language using Bidirectional Long Short
Term Memory (BiLSTM), eliminating the need for most feature construction.
Entity contains people, location, grouping, date_time_month, numerical
values, etc. Myanmar expression is still ambitious to analyze Name Entity
(NE) as well as familiar conversation so it bags of geographical instruction
towards noticeable items, never barrier explanation among words and none
capitalization comparable other languages. Myanmar Natural Language
Processing (NLP) is told to be closed growing along with has directly been
excruciating to be matured. Considering that logic, Entity Identification (EI)
entitled collection for Burma ER analysis is annually explained and built as
composing that monograph. The elucidate EI bulk is crucial for Myanmar
ER research’s improvement . For planned entity classification research, those
entity titled compilation is tested all the while entire the aimed evidence for
Burma ER and it will also be determined. By using BiLSTM based network
architecture, the best accuracy is achieved with 83.62%. Accordingly, here
task dispose of the aspect engineering development and does not demand to
acquire not only expression but also territory ability.
2023-01-01T00:00:00Z