CUSTOMER SENTIMENT ANALYSIS BASED ON LATENT DIRICHLET

ALLOCATION (LDA) TECHNIQUE

ABSTRACT

With the rapid development of e-commerce, most customers express their opinions on various
kinds of entities, such as products and services. Reviews generally involves specific product
feature along with opinion sentence. These reviews have rich source of information for decision
making and sentiment analysis. Sentiment analysis refers to a classification problem where the
main focus is to predict the polarity of words and then classify them into positive, negative and
neutral feelings with the aim of identifying attitude and opinions. This paper describe Latent
Dirichlet Markov Allocation Model (LDMA), a new generative probabilistic topic model, based
on Latent Dirichlet Alloction (LDA) and Hidden Markov Model (HMM), which emphasizes on
extracting topics from consumer reviews. After the topic extraction, use SentiWordNet
dictionary for sentiment classification. Experimental results show that the proposed technique
overcomes the previous limitations and achieves higher accuracy when compared to similar
techniques.

1. INTRODUCTION

Emotion expression plays a vital role in various part of every-day communication. In past,
various measures have been used toevaluate it, through a combination of indications such as
facialexpressions, gestures, and actions etc. Emotions extraction usingfacial, gestures and action
are the part of digital image processingand computer vision. Emotions extraction is more
difficult fromtexts especially from multi-languages texts, like in posts on socialmedia and
customers’ reviews. [1] This type of data has presence ofambiguity and complexity of words in
terms of meaning makethem more difficult. Factors such as users writing style, politeness, irony,
variability in language is one of the important problems inextraction of emotions.[2] A wide
variety of state-of-art work hasbeen carried out in the domain of opinions mining and
sentimentanalysis but limited research are focused on detection/extractionof emotions in tweeter.
Cluster analysis or clustering is the task of grouping a set of objects in such a way that objects in
the same group (called a cluster) are more similar (in some sense or another) to each other than
to those in other groups (clusters). [3] It is a main task of exploratory data mining, and a common
technique for statistical data analysis, used in many fields, including machine learning, pattern
recognition, image analysis, information retrieval, bioinformatics, data compression,
and computer graphics. [4] PSMs contain inference actions which need specific knowledge in
order to perform their task. For instance, Heuristic Classification needs a hierarchically
structured model of observables and solutions for the inference actions abstract and refine,
respectively. So a PSM may be used as a guideline to acquire static domain knowledge.

• A PSM allows to describe the main rationale of the reasoning process of a KBS which
supports the validation of the KBS, because the expert is able to understand the problem solving
process. In addition, this abstract description may be used during the problem solving process
itself for explanation facilities.

Cluster analysis itself is not one specific algorithm, but the general task to be solved.[5] It can be
achieved by various algorithms that differ significantly in their notion of what constitutes a
cluster and how to efficiently find them. Popular notions of clusters include groups with
small distances among the cluster members, dense areas of the data space, intervals or
particular statistical distributions. Clustering can therefore be formulated as a multi-objective
optimization problem. [6] The appropriate clustering algorithm and parameter settings (including
values such as the distance function to use, a density threshold or the number of expected
clusters) depend on the individual data set and intended use of the results. Cluster analysis as
such is not an automatic task, but an iterative process of knowledge discovery or interactive
multi-objective optimization that involves trial and failure. It is often necessary to modify data
pre-processing and model parameters until the result achieves the desired properties.

2. LITERATURE REVIEW
Most papers suggest using an existing conventional clustering algorithm (e.g., weighted
k-means in CluStream) where the micro-clusters are used as pseudo points. Another approach
used in Den Stream is to use reach ability where all micro-clusters which are less than a given
distance from each other are linked together to form clusters. Grid-based algorithms typically
merge adjacent dense grid cells to form larger clusters (see, e.g., the original version of D-Stream
and MR-Stream).

Clustering Performance on Evolving Data Streams: Assessing Algorithms and Evaluation

Measures within MOA

Author - Philipp Kranen ; Hardy Kremer ; Timm Jansen ; Thomas Seidl

In today's applications, evolving data streams are ubiquitous. Stream clustering algorithms were
introduced to gain useful knowledge from these streams in real-time. The quality of the obtained
clusterings, i.e. how good they reflect the data, can be assessed by evaluation measures. A
multitude of stream clustering algorithms and evaluation measures for clusterings were
introduced in the literature, however, until now there is no general tool for a direct comparison of
the different algorithms or the evaluation measures. In our demo, we present a novel
experimental framework for both tasks. It offers the means for extensive evaluation and
visualization and is an extension of the Massive Online Analysis (MOA) software environment
released under the GNU GPL License.

Organizing multimedia big data using semantic based video content extraction technique

Author - Manju ; P. Valarmathie

With the proliferation of the internet, video has become the principal source. Video big data
introduce many hi-tech challenges, which include storage space, broadcast, compression,
analysis, and identification. The increase in multimedia resources has brought an urgent need to
develop intelligent methods to process and organize them. The combination between multimedia
resources and Semantic link Network provides a new prospect for organizing them with their
semantics. The tags and surrounding texts of multimedia resources are used to measure their
association relation. There are two evaluation methods namely clustering and retrieval are used
to measure the semantic relatedness between images accurately and robustly. This method is
effective on image searching task. The semantic gap between semantics and video visual
appearance is still a challenge. A model for generating the association between video resources
using Semantic Link Network model is proposed. The user can select the attributes or concepts
as the search query. This is done by providing the knowledge conduction during information
extraction and by applying fuzzy reasoning. The first action line is related to the establishment of
techniques for the dynamic management of video analysis based on the knowledge gathered in
the semantic network. This helps the decisions taken during the analysis process. Based on a set
of rules it is able to handle the fuzziness of the annotations provided by the analysis modules
gathered in the semantic network.

Evaluation Methodology for Multiclass Novelty Detection Algorithms

Author - Elaine R. Faria ; Isabel J. C. R. Goncalves ; Joao Gama

Novelty detection is a useful ability for learning systems, especially in data stream scenarios,
where new concepts can appear, known concepts can disappear and concepts can evolve over
time. There are several studies in the literature investigating the use of machine learning
classification techniques for novelty detection in data streams. However, there is no consensus
regarding how to evaluate the performance of these techniques, particular for multiclass
problems. In this study, we propose a new evaluation approach for multiclass data streams
novelty detection problems. This approach is able to deal with: i) multiclass problems, ii)
confusion matrix with a column representing the unknown examples, iii) confusion matrix that
increases over time, iv) unsupervised learning, that generates novelties without an association
with the problem classes and v) representation of the evaluation measures over time. We
evaluate the performance of the proposed approach by known novelty detection algorithms with
artificial and real data sets.

Performance evaluation of distance measures for preprocessing of set-valued data in

feature vector generated from LOD datasets

Author - Rajesh Mahule ; AkshenndraGarg

The linked open data cloud has evolved as a huge repository of data with data from various
domains. A lot of work has been done in generating these datasets and enhancing the LOD cloud,
whereas a little work is being done in the consumption of the available data from the LOD. There
are several types of applications that have been developed using the data from the LOD cloud; of
which, one of the areas that has attracted the researchers and developers most is the use of these
data for machine learning and knowledge discovery. Using the available, state of the art
knowledge discovery and machine learning algorithms requires conversion of the heterogeneous
interlinked RDF graph datasets, available in LOD cloud, to a feature vector. This conversion is
performed with the subject set as instances; the predicates set as attributes and object set as
attribute values in a feature vector However, choosing the most suitable distance measures of the
different distance measures available is a problem that needs to be catered. This paper provides a
performance study to select the most suitable distance measure that can be used in pre-processing
by building the feature vector with the different distance measures for set-valued data attributes
and applying transformation with Fastmap. The evaluation of the distance measures is done
using clustering of the transformed feature vector table with pre-identified class labels and
getting micro-precision values for the clustering results. Performing the experimental analysis
with LMDB data it has been found that the Hausdorff and RIBL distance measures are the most
suitable distance measures that can be used to pre-process the created feature vector with set-
valued data from the linked open data cloud.

Analyzing Enterprise Storage Workloads With Graph Modeling and Clustering

Author - Yang Zhou ; Ling Liu ; SangeethaSeshadri

Utilizing graph analysis models and algorithms to exploit complex interactions over a network of
entities is emerging as an attractive network analytic technology. In this paper, we show that
traditional column or row-based trace analysis may not be effective in deriving deep insights
hidden in the storage traces collected over complex storage applications, such as complex spatial
and temporal patterns, hotspots and their movement patterns. We propose a novel graph analytics
framework, GraphLens, for mining and analyzing real world storage traces with three unique
features. First, we model storage traces as heterogeneous trace graphs in order to capture
multiple complex and heterogeneous factors, such as diverse spatial/temporal access information
and their relationships, into a unified analytic framework. Second, we employ and develop an
innovative graph clustering method that employs two levels of clustering abstractions on storage
trace analysis. We discover interesting spatial access patterns and identify important temporal
correlations among spatial access patterns. This enables us to better characterize important
hotspots and understand hotspot movement patterns. Third, at each level of abstraction, we
design a unified weighted similarity measure through an iterative dynamic weight learning
algorithm. With an optimal weight assignment scheme, we can efficiently combine the
correlation information for each type of storage access patterns, such as random versus
sequential, read versus write, to identify interesting spatial/temporal correlations hidden in the
traces.

3. PROPOSED METHODOLOGY

This proposed Navie bayes model that classifies documents into reader-emotion categories. they
studied the classification of news articles into different sentiment classes representing the
emotions they trigger in their readers. This work mainly differs from other literature in focusing
more on what the reader would feel while reading the article rather than what the writer was
feeling while writing it. Other than the classification itself, which has been detailed in our
previous work, we study the impact of the number of sentiment classes on the classification
performance (i.e., accuracy, precision, and recall). It analyze the results of the different
experiments and conclude with the limitations that make multi-class classification a difficult
task.
ADVANTAGES

 LDA is a powerful topic modeling technique that can uncover latent topics within a
collection of text documents.
 LDA enables granular analysis by breaking down customer feedback into specific topics
or themes. This allows businesses to gain insights into the various aspects of their
products, services, or brand that are positively or negatively perceived by customers.
 With the right implementation and infrastructure, LDA-based sentiment analysis systems
can provide real-time insights into customer sentiment.
 By identifying specific topics or themes driving customer sentiment, businesses can
derive actionable insights to improve products, services, marketing strategies, and
customer experience.
 Fig.1 WORKFLOW OF PROPOSED METHOD.

4. RESULTS AND DISCUSSIONS

Data cleaning is one of the most important processes for obtaining accurate experimental
results. To test our model, we used the IMDB movie review dataset . The dataset
included 50,000 reviews that were evenly divided into positive and negative reviews. The
dataset was divided into a 80% (4000) training set and a 20% (1000) testing set. First,
unnecessary columns were removed. The second process included some spelling
corrections, removing weird spaces in the text, html tags, square brackets, and special
characters represented in text and contraction. They were then handled with emoji by
converting them to the appropriate meaning of their occurrence in the document.
Thereafter, we put all the text to lowercase, and removed text in square brackets, links,
punctuation, and words containing numbers. Next, we removed stop words because
having these makes our analysis less effective and confuses our algorithm. Subsequently,
to reduce the vocabulary size and overcome the issue of data sparseness, stemming,
lemmatization, and tokenization processes were applied. We normalized the text in the
dataset to transform the text into a single canonical form. With the aim of achieving
better document classification, we also performed count vectorization for the bag-of-
words (BOW) model. The BOW model can be used to calculate various measures to
characterize the text. For this calculation process, the term frequency-inverse document
frequency (TF-IDF) is the best method. Basically, TF-IDF reflects the importance of a
word. We applied the N-gram model to avoid the shortcomings of BOW when dealing
with several sentences with words of the same meaning. The N-gram model parses the
text into units, including TF-IDF values. The N-gram model is an effective model
representation used in sentiment analysis. Each N-gram related to parsed text becomes an
entry in the feature vector with the corresponding feature value of TF-IDF. After
preprocessing, the LDA model was applied. LDA is a three-level hierarchical Bayesian
model that creates probabilities at the word level, on the document level, and on the
corpus level
 MOST COMMON SENTIMENTS
,413
49 film good man watch
,464 see get act
43
,471
,373 26
,690,279,762,783,588 20 ,428,583,096,026,160
14 16 12 14 12 14 12 15 13 15

POSITIVE NEGATIVE

Figure 2. Most Common Sentiments.

5. CONCLUSION
In this project was studied the task of multi-class sentiment analysis and evaluated the
evolution of various KPIs as the number of sentiment classes increased. It analyzed the
difficulties of, and the different challenges involved with, multi-class classification, and
proposed some metrics to measure the distance between sentiments (i.e., how similar they
are to one another)and concluded that even though the task of multi-class analysis is
important, it might be more interesting to perform a sentiment detection task through
which all of the sentiments present within a text are extracted. In future this work will be
experimented and tested in the public cloud based peta-sized datasets.

6. REFERENCES
1. S. Guha, N. Mishra, R. Motwani, and L. O’Callaghan, “Clustering data streams,”
in Proceedings of the ACM Symposium on Foundations of Computer Science, 12-
14 Nov. 2000, pp. 359–366.
2. C. Aggarwal, Data Streams: Models and Algorithms, ser. Advances in Database
Systems, Springer, Ed., 2007.
3. J. Gama, Knowledge Discovery from Data Streams, 1st ed. Chapman &
Hall/CRC, 2010.
4. J. A. Silva, E. R. Faria, R. C. Barros, E. R. Hruschka, A. C. P. L. F. d. Carvalho,
and J. a. Gama, “Data stream clustering: A survey,” ACM Computing Surveys,
vol. 46, no. 1, pp. 13:1–13:31, Jul. 2013.
5. C. C. Aggarwal, J. Han, J. Wang, and P. S. Yu, “A framework for clustering
evolving data streams,” in Proceedings of the International Conference on Very
Large Data Bases (VLDB ’03), 2003, pp. 81–92.
6. F. Cao, M. Ester, W. Qian, and A. Zhou, “Density-based clustering over an
evolving data stream with noise,” in Proceedings of the 2006 SIAM International
Conference on Data Mining. SIAM, 2006, pp. 328–339.
7. Y. Chen and L. Tu, “Density-based clustering for real-time stream data,” in
Proceedings of the 13th ACM SIGKDD International Conference on Knowledge
Discovery and Data Mining. New York, NY, USA: ACM, 2007, pp. 133–142.
8. L. Wan, W. K. Ng, X. H. Dang, P. S. Yu, and K. Zhang, “Density based clustering
of data streams at multiple resolutions,” ACM Transactions on Knowledge
Discovery from Data, vol. 3, no. 3, pp. 1–28, 2009.
9. L. Tu and Y. Chen, “Stream data clustering based on grid density and attraction,”
ACM Transactions on Knowledge Discovery from Data, vol. 3, no. 3, pp. 1–27,
2009.
10. M. Ester, H.-P.Kriegel, J. Sander, and X. Xu, “A density-based algorithm for
discovering clusters in large spatial databases with noise,” in Proceedings of the
ACM SIGKDD International Conference on Knowledge Discovery and Data
Mining (KDD’1996), 1996, pp. 226–231.