Sei sulla pagina 1di 7

EMS Engineering Science Journal

Study Article

Data Mining of Social Media Statistical Techniques for


Analyzing Student’s Problems in Universities

Pridhivi Raj Kandula1,*


1
Business Analytics and Project Management, University of Connecticut, USA

*
Corresponding author : Mr. Pridhivi Raj Kan- Abstract
dula, Business Analytics and Project Manage-
ment, University of Connecticut, Unit: UCo- Online networking permits the creation and associations of client made substance.
nn-Hartford, 100 Constitution Plaza Hartford, Social medium spots incorporate Facebook, Twitter and so on. Understudy's easy go-
CT 06103, USA ,Tel: +1(860) 728-2451,Email: ing talk via web-based networking media centered into their instructive experience,
msbapm@business.uconn.edu mentality, and stress over the learning method. Data from such un-instrumented
situations can exhibit important information to report understudy issue. Inspect-
Received: 05-03-2018
ing information from such a social media is testing undertaking. The issue of under
Accepted: 26-03-2018
study's encounters uncovers from web-based social networking sited require human
Published: 31-03-2018
investigation or Interaction. It focuses on building understudy's Twitter presents on
Copyright: © 2018 Pridhivi Raj Kandula
known issue and inconveniences in their instructive practices. This paper proposes
a work process to assemble both subjective examination and huge scale information
mining plan. Initial an example is taken from understudy and afterward subjective
examination led on that example which is related to building understudy's instructive
life. So just tweets identified with designing understudy is gathered. It is discovered
that building understudies experience issues, for example, substantial learning load,
absence of social gathering, and rest insufficiency. In view of this result, a multi-name
arrangement calculation that is Naive Bayes Multi-Name Classifier calculation is con-
nected to sort tweets displaying understudy's issues. At that point choice tee calcu-
lation is connected to make more precise outcome it will perform separating. The
calculation readies an identifier of understudy issues. This investigation exhibits a
strategy and result that show how easygoing web-based social networking informa-
tion can introduce understanding into understudy's occurrence.

Keywords: Social Networking; Web-Text Analysis; Education; Social Network Anal-


ysis; Computer and Education, Data Mining

Introduction
The procedure of gathering of information, looking and ex- to bottom and wonderful comprehension of understudies’
amining a gigantic measure of information in a database, in encounters particularly in their learning-related issues and
order to discover examples or connections is the utilization troubles. To decide in a tweet, what are the understudy’s con-
of information mining to recognize extortion. In fact, infor- cern, is a more convoluted errand than to decide the feeling of
mation mining is the procedure to discover connections or a tweet notwithstanding for a human judge. Along these lines,
examples among many fields in extensive social databases. this investigation requires a subjective examination, and is
Robotized forecast of floats and practices: Mining mechaniz- difficult to do in a completely unendorsed way. Estimation ex-
es the way toward seeking prescient data in an expansive da- amination is, hence, not pertinent to this investigation. In this
tabase [1]. Inquiries that regularly require broad hands-on investigation, we will actualize a multi-name grouping model
investigation would now be able to be effortlessly replied where we enable one tweet to fall into various classifications
from the information. An average case of a prognostic issue is in the meantime. Our work broadens the extent of informa-
focused on advertising. Our motivation is to accomplish top tion driven methodologies in instruction, for example, learn-
Cite this article: Pridhivi Raj Kandula. Data Mining of Social Media Statistical Techniques for Analyzing Student’s Problems in Universities.
EMS Eng Sci j 2018, 2(1):009.
ing investigation and instructive information mining. Gener- Qualitative Investigation
ally, instructive analysts are utilizing techniques, for example,
overviews, interviews, center gatherings, classroom exercis- Qualitative investigation is a method of examination utilized
es to gather information identified with understudies’ learn- in numerous differing scholarly control, by convention in the
ing encounters. These techniques are normally exceptionally sociologies, yet in addition in statistical surveying and further
tedious, so can’t be cloned or rehashed with high recurrence. settings. Subjective analysts intend to accumulate an inside
and out comprehension of human activities and the reasons
The exploration objective of this learning is: that oversee such conduct. The subjective strategy analyzes
the why and how of basic leadership, not exactly what, where,
 To demonstrate a work of online networking data when. Subsequently, minor however engaged examples are
sense-production for instructive reasons, joining both frequently utilized than tremendous examples. Subjective
subjective examination and extensive scale information systems make data just on the specific cases contemplated,
mining strategies. and any broader terminations are just proposals. Quantita-
 To find designing understudy’s easygoing exchang- tive strategies would then be able to be utilized to search for
es on Twitter, with a specific end goal to knows the issue trial bolster for such research hypotheses.
coming into their life. This Study wants to center around
building understudy’s remarks posted on Twitter about Related Work
their issues in composition life on the grounds that:
The hypothetical reason for the estimation of casual informa-
1. Designing schools and branch have for some time been tion on the web can be drawn from Goffman’s hypothesis of
worried with understudy business and conservation themes. social execution [3]. Goffman’s hypothesis of social execution
Designing graduates contain a noteworthy piece of the coun- is broadly used to give detail of intervened cooperation on the
try’s potential work compel and directly affect the country’s web today [4]. One of the most key parts of this hypothesis is
money related development [2]. the thought of front-arrange and back-phase of individuals’
2. In view of comprehension of understudies trouble leaders social exhibitions. Contrasted and the front stage, the un-
can make more proficient conclusions on legitimate obstruc- winding air of back-arrange typically acclaims more uncon-
tion that can help understudies to overcome snags in training strained activities. For understudies, contrasted with formal
and help the understudy to take care of the issue. classroom settings, online networking is relative casual and
3. Twitter is an all-around loved online networking website. unwinding back-arrange. At the point when understudies
Its substance is as often as possible open and extremely short post content via web-based networking media locales, they
that is close to 140 characters for each tweet. Twitter offer as a rule post what they think and feel right then and there.
free APIs that is utilized to stream information and enables In this sense, the information gathered from online discus-
designers to expand upon and broaden their applications in sion might be more bona fide and unfiltered than reactions to
new and imaginative ways. Access Data from Twitter give de- formal research prompts. Numerous investigations demon-
signers low idleness access to Twitter’s worldwide stream of strate that web-based social networking clients may deliber-
Tweet information. To develop an information mining plan or ately deal with their online character to “look better “than, in
are associated with investigation look into, the Streaming API actuality, [5,6] .Other examinations demonstrate that there is
is most appropriate for such things. Twitter realities are in a shortage of mindfulness about overseeing on the web char-
reasonable arrangement for examination. Twitter’s terms of acter among understudies [7], and that youngsters typical-
apply for the information are generally tolerant. It is by and ly view web-based social networking as their own space to
large acknowledged that tweets are open and accessible to hang out with peers outside seeing guardians and instructors
anyone; henceforth they allow entrée to any record with no [8]. Understudies’ online discussions uncover parts of their
compelling reason to ask for authorize. encounters that are not effectively found in formal classroom
settings, along these lines are typically not archived in in-
Social Networks Analysis structive writing. The wealth of online networking informa-
tion gives openings yet additionally displays methodological
Social Networks Analysis (SNA), or basic examination, goes challenges for dissecting huge scale casual printed informa-
for considering connections between people, rather than sin- tion. The following area audits famous strategies utilized for
gular characteristics or properties. An interpersonal organi- breaking down Twitter information. Gaffney [9] examina-
zation is thought to be a gathering of individuals, an associa- tions tweets with hashtag #iranElection utilizing histograms,
tion or social people who are associated by social connections client systems, and frequencies of best watchwords to gauge
like companionship, agreeable relations, or enlightening online activism. Comparative examinations have been led in
trade. Distinctive DM strategies have been utilized to mine in- different fields including social insurance [6], showcasing [5],
formal communities in instructive conditions, yet synergistic and sports [3]. These investigations have more accentuation
arranging is the most well-known. Synergistic separating or on factual models and calculations. They cover an exten-
social sifting is a strategy for making programmed expecta- sive variety of subject’s fame expectation, occasion location,
tions about the interests of a client by gathering taste inclina- theme revelation and tweet order. Conclusion investigation
tions from numerous clients. is exceptionally helpful for mining client assessments on
items or organizations through their surveys or online posts.
It finds wide reception in advertising and client relationship
Cite this article: Pridhivi Raj Kandula. Data Mining of Social Media Statistical Techniques for Analyzing Student’s Problems in Universities.
EMS Eng Sci j 2018, 2(1):009.
2/7
administration (CRM). commotion in content. So, we preprocessed the writings be-
fore preparing the classifier.
Different examinations demonstrate that there is an absence
of mindfulness about overseeing on the web character among 1. First evacuate each designing Problems hash labels. What’s
understudies, and that youngsters for the most part view web- more, for new happening hash labels, simply evacuated the
based social networking as their own space to hang out with sign, and held the hash label writings.
peers outside seeing guardians and instructors. Understudy’s 2. For recognizing negative feeling and issues negative words
online discussions uncover parts of their encounters that are are utilized. in this way it supplants words completing with
not effectively found in formal classroom settings, in this way “n’t” and additionally visit negative words (e.g. actually no,
are normally not archived in instructive writing. The follow- not, nothing) as negtoken”.
ing are surveys of concentrates on Twitter from the fields of 3. Detached each single word that incorporates non-letter
information mining, machine learning, and regular dialect images and accentuation. This fused the cancellation of @
preparing. These investigations have more accentuation on and http joins. Additionally, erase all the RTs.
measurable models and calculations. They cover an extensive 4. For recreated letters inside words, arrangement when it
variety of themes prevalence forecast, occasion recognition, finds two coordinating letters duplicating, it held them two.
point disclosure and tweet grouping. Among these subjects, On the off chance that it distinguished in excess of two same
tweet grouping is most applicable to this examination. Well Letters reproducing, substitute them with one letter. Subse-
known arrangement calculations incorporate Naïve Bayes, quently, “soooocuuuteeee” is remedied to “”So charming”. At
Decision Tree, Logistic Regression, Maximum Entropy, Boost- first exact words, for example, “Sweet” and “pal” were kept as
ing, and Support Vector Machines (SVM). they were.
5. At this point it utilized the Lemur information recupera-
Most existing examinations found on tweet grouping are ei- tion toolbox to dispose of the continuous stop words. It kept
ther twofold order on pertinent and unimportant substance, words like “much, parcel, some, all, eternity, still, just”, on the
or multi-class characterization on non-specific classes, for ex- grounds that the tweets.
ample, news, occasions, feelings, arrangements, and private
messages. Supposition examination is another exceptionally 2 Naive Bayes Multi-Label Classifier
well known three-class arrangement on positive, negative,
or unbiased feelings/sentiments. Assessment examination is The Naive Bayes classifier is a straightforward probabilistic
exceptionally helpful for mining client sentiments on items classifier which is based on Bayes theorem with strong and
or organizations through their audits or online posts. It finds naïve self-government assumptions. It is one of the most
wide selection in showcasing and client relationship admin- basic text categorization method with various applications
istration (CRM). Numerous techniques have been produced in email spam exposure, private mail sorting, document
to mine opinion from writings our motivation is to accom- categorization, language discovery and sentiment discov-
plish further and better comprehension of understudy’s en- ery. Naive Bayes executes well in many difficult real-world
counters particularly their learning-related issues and issues. troubles. Even though it is frequently outperformed by oth-
To figure out what understudy issues a tweet shows is a more er techniques such as boosted trees, Max Entropy, Support
entangled undertaking than to decide the assumption of a Vector Machines etc., Naive Bayes classifier is extremely effi-
tweet notwithstanding for a human judge. Hence, our investi- cient since it is less computationally, and it requires a small
gation requires a subjective examination. amount of preparation information. One well-liked way to
execute multi-label classifier is to convert the multi-label or-
Algorithms Used ganization problem into multiple single-label categorization
problems.
This examination constructed a multilabel classifier to sort
tweets remains on the classifications created in content in- Next is the necessary action of the multi-label Naive Bayes
vestigation stage. There are various very much preferred classifier. Assume there are sum of W words in the prepar-
classifiers for the most part utilized as a part of information ing document compilation in this case, every tweet is a doc-
mining and machine learning field. It builds up that Nave ument.
Bayes classifier to be exceptionally proficient for this dataset
contrasted and assist multilabel classifiers [7]. D = d1;d2; : : : ; dw, and a total amount of M categories K =
k1; k2; : : : ; kM. If a word dw appears in a category k for n d
1 Text Pre-Processing wk times, and appear in categories other than k for n d wk ′
times, then, the probability of this word in a definite category
Twitter customer utilizes different strange images to express c is
certain centrality. For Ex, is utilized to determine a hash tag, @
used to indicate a client record, and RT is utilized to demon-
strate a retweet. Twitter clients periodically copy letters in
words in this way to feature the words, for ex, “soooocuuu-
teeee”, “verrrryyyysmmmaaaart”, and “Looooking amazing”. Similarly, the probability of this word in categories other than
Furthermore, basic stop words, for example, “an, an, and, of, c is:
he, she, it”, nonletter images, and accentuation also convey
Cite this article: Pridhivi Raj Kandula. Data Mining of Social Media Statistical Techniques for Analyzing Student’s Problems in Universities.
EMS Eng Sci j 2018, 2(1):009.
3/7
a decision tree. It builds up a decision tree for a given train-
ing set T made up of set of training instances. Training set is
a group of records. Each record contains a set of attributes;
one of the attributes is the class. An instance is defined as val-
Suppose there are an entire number of X documents in the ues of attributes and a class. Decision trees classify instanc-
preparing set, and K of them are in category k. Then the prob- es by opening at the root of the tree and moving all the way
ability of category k is through it until a leaf node. Let the classes be represented
by {C1, C2, …,Cn}. Originally, the class occurrence is comput-
ed for instance in training set T. If all instances are belonging
And the probability of other categories k’ is to the similar class, node K with that class is build. However,
if set T include instances that is belonging to more than one
class, then the test for choosing attribute for dividing is carry
out and the attribute fulfilling dividing criteria is chosen for
the test at that node. The training set T is then divided into k
For a document di in the trying set, there are Y words Wdi =
limited subsets {T1, T2, …,Tk } on the base of this assessment
wi1;wi2; : : : ; wiY, and Wdi is a subset of D. The purpose is to
and the algorithm is recursively applied on every nonempty
classify this document into category c or not c. We assume
division. The algorithm for creation of a decision tree is given
independence among each word in this document, and any
below.
word wik conditioned on k or k’ follows multinomial distri-
bution. Therefore, according to Bayes Theorem, the probabil-
1) Create (T)
ity that di fit in to category k is
2) Computefreq (Ci, T).
3) I fall instances fit in to similar class, then return leaf.
4) For each attribute attesting for splitting criteria is con-
ducted Attribute fulfilling test is test node K.
and the probability that di fit into group other than c is 5) Recursively create (Ti) on each division Ti. Insert those
nodes as child of node K.
6) Stop.

Choosing the best attribute for splitting


Because p k di + p(k ′ + di),it normalize the latter two items
which are comparative to p k di and p k′ di to get the actual The selection of attribute at each node is a most important
values of p k di . If p k di is larger than the probability thresh- procedure. Various approach projected to select the best at-
old T, then di fit into category k, otherwise, di does fit into cat- tribute.
egory k. Then do again this process for every category. In this
execution, if for a definite document, there is none category Using Information Theory
with a positive probability larger than T, it allocates the one
category with the largest probability to this document. In cal- The information gain is depending on the reduce in entropy
culation, others is an special group. A tweet is only allocated behind a dataset is divide on an attribute. creating a decision
to others while others is the just category among probability tree is all concerning the discovering attribute that returns
greater than T. the maximum information gain. Purpose of process of data
categorization is to exploit the information gain as it leads
1. The Tree Construction Algorithm to rise in categorization accuracy. Information gain is an im-
purity-based criterion that uses the entropy measure (origin
Decision trees are influential and well-liked tools for catego- from information theory) as the impurity measure. The gain
rization. Decision tree is a classifier which is in the type of is described as the information get from a message based on
a tree organization. The highest choice node in a tree which its probability P. For any set of instances T, the probability
matches to the best interpreter named root node. that an instance fit in to class Ci is specified as
P = frequency (Ci,T) |T| (1)
Decision node: specifies a test on attribute and have two or
more branches. Where |T| is number of instances in set T and freq (Ci,T) rep-
resented the number of instances in T that fit in to class Ci..
Leaf node: It indicates the value of the target attribute i.e. rep- Now, the average information enclosed in set T concerning
resents a classification or decision. class relationship of instances, called entropy of set T, and is
calculated in bits as in for T = Pi k i=1 × log2(Pi)bits(2)
Arc/edge: Split of one attributes
k is the number of classes in set T. The test X performed at a
Path: A disjunction of examination to make the final conclu- node on the chosen attribute gives subsets T1, T2, …,Tk . The
sion [3]. information by this dividing process is calculated as the sum
over these subsets, is given as in for x T |Ti| |T| × k i=1 in for
This algorithm uses a divide and conquers approach to build Ti (3)
Cite this article: Pridhivi Raj Kandula. Data Mining of Social Media Statistical Techniques for Analyzing Student’s Problems in Universities.
EMS Eng Sci j 2018, 2(1):009.
4/7
which is related to building understudies instructive life [4].
The reduction in entropy due to portioning of T with test X on It discovered designing understudies experience issues, for
the chosen attribute, represented as Gain (X), is calculated as example, overwhelming learning load, absence of social gath-
gain X = in for T – in for x (T) (4) ering, and rest lack. remain on these results, creators apply
a multi-mark characterization calculation to classify tweets
The attribute which gives highest information gain is pre- introducing understudy’s issues. At that point choice tee cal-
ferred. The trouble with above approach for accumulation culation is connected to make more precise outcome it will
of test quality at a hub is that it is one-sided towards char- perform separating after that utilized the calculation to set
acteristics with a considerable measure of qualities when up a finder of understudy issues. This examination displays
contrasted with traits with less esteems and it prompts ex- a strategy and result that exhibit how easygoing web-based
pansive choice trees that feebly sum up the trouble. This is- social networking information can show understanding into
sue can be evacuated with standardization of pick up basis understudy’s occurrence.
and use of pick up proportion. The pickup proportion figures
proportion of data produced by partioning T and is commu- In this examination it actualized a multi-name characteri-
nicated as zation display where we allowed one tweet to go down into
numerous classifications in the meantime. Our arrangement
is contrasted and other non-specific groupings. Our work
(5) grows the scope of information driven methodologies in in-
structing, for example, learning examination and instructive
The split info(X) calculates information gained by splitting information mining.
training set T into k subsets on test X.
The critical point in proposed think about is, first, it propos-
es a work process to connect and coordinate a subjective
(6) research philosophy and vast scale information mining sys-
tems. It constructs our information mining calculation with
The characteristic is decided for the test which acquires respect to subjective knowledge coming about because of
greatest pick up proportion. This approach has issue that it human seeing, so it can increase further comprehension of
tends to support properties for which split info(X) is little. the information. At that point apply the calculation to anoth-
Another issue is that pick-up proportion can be ascertained er substantial scale and unexplored dataset, with the goal
just when the split info(X) is nonzero. To overcome this issue, that the physical strategy is made strides. Second, the paper
Quinlan prescribed figure data increase over each one quali- gives profound bits of knowledge into building understudy’s
ties and considering characteristics with data pick up which instructive encounters as responded in casual, uncontrolled
is in any event as substantial as normal of data increase over situations. Numerous issues and issues, for example, consid-
all properties. The utilization of pick up proportion gives en- er life adjust, absence of rest, absence of social engagement,
hanced precision and intricacy of classifier. and absence of assorted variety obviously develop. These
could bring attention to instructive specialist, arrangement
Data pick up is typically a decent quality measure for choos- producer.
ing the pertinence of a property, it isn’t perfect. A critical
trouble emerges when data pick up is identify with qualities System Architecture
that can go up against an immense number of partitioned es-
teems. For instance, accept that one is building a choice tree
for a few information concerning the clients of a business.
Data pick up is much of the time apply to settle on a choice
that which of the traits are the primarily proper, so they can
be tried near the base of the tree. Data pick up proportion
is utilized now and then. This predisposition the choice tree
contrary to trusting qualities with countless esteems. Not-
withstanding, qualities with exceptionally little data esteems
at that point wind up noticeable to acquire an unjustifiable
advantage.

Proposed Method

Proposed framework built up a work process to assemble


both subjective examination and vast scale information min-
ing systems it focuses on building understudy’s Twitter pres-
ents on known issue and inconveniences in their instructive
Fig 1: System workflow
practices. Initial an example is taken from understudy and
after that it direct subjective examination on that example

Cite this article: Pridhivi Raj Kandula. Data Mining of Social Media Statistical Techniques for Analyzing Student’s Problems in Universities.
EMS Eng Sci j 2018, 2(1):009.
5/7
System Flow Extension of Categories

The framework stream is appeared in the figure underneath: The 5 conspicuous subjects are: substantial examination
In this framework there is an investigative strategy to locate stack, absence of social engagement, negative feeling, rest
the fitting information and suitable Twitter hash labels, and issues, and decent variety issues. Each subject uncovers one
afterward a Twitter hash tag is a word that is beginning with issue or troubles that building understudies have in their life.
a sign, which is utilized to feature or label an issue. It sets up that numerous no of tweets fit in to in excess of one
classification. For instance, “I am nourished up of study why
Social event of tweets utilizing the hash label building Prob- I’m not in move school? Loathe being in designing school. An
lems. This relates to the stage 1 In Fig. 1. excess of stuff. Far excessively troublesome. No appreciate”
comes into substantial examination load, and negative feeling
Next in the stage 2 and 3 of Fig. 1. The inductive substance in the meantime. Henceforth one tweet can have numerous
investigation is performing on the example on designing is- classifications. This is a multilabel grouping as complexity to
sue. In stage 4, it is discovered that the significant issue that a solitary mark arrangement in which each tweet can fall just
comes into designing understudies fall into various surely in one class. The no of classifications where one tweet fit in to
understood classes. In view of these classifications, a mul- are called tweets names generally name set.
tilabel Nave Bayes arrangement calculation and choice tree
classifier is executed for characterization [10]. The Prominent Themes

In stage 5 the execution of the classifiers is assessed by con- If the 1 tweet comes into numerous classes, it is checked a
trasting it and other multilabel classifiers. few occasions. Measure of tweets in each gathering look at.
Here extensive measure of tweets falls into “Others”. If it’s
In stage 6 the grouping calculation is connected by System to not too much trouble note, model tweets displayed in each
set up a locator that assistance acknowledgment of building subject may likewise go down in various further classes at a
understudy’s issues. The outcomes are given by stage 7 help similar example, aside from ones in “others.
teachers to recognize in danger understudies and settle on
choices on legitimate obstruction to save them. Conclusion
Mining online networking information is useful to analysts
in learning investigation, instructive information evacuation,
and learning expertise. It gives an approach to looking at so-
cial medium insights that overcome the principle limitations
of both physical subjective investigation and tremendous
scale computational investigation of client delivered printed
content. Two calculations are helpful for this characterization
first the Naive Bayes Multi-Name Classifier and Second the
Tree development calculation. It tells instructive director, and
other pertinent appraisal creators to grow additionally toler-
ating of designing understudies’ foundation understanding.

References
1. Xin Chen, MihaelaVorvoreanu, Krishna Madhavan. Mining
Social Media Data for Understanding Students’ Learning Ex-
periences. IEEE tarnsactions on learning Technologies. 2014,
7(4): 246-259.

2. Rost M, Barkhuus L, Cramer H, Brown B. Representation


Fig.2. Architecture and communication: challengesin interpreting large social
media datasets. In Proceedings of the 2013 conference on
Inductive Content Analysis Computer Supported cooperative work. 2013, 357–362.

There are none predefined classifications of the information 3. Clark M, Sheppard S, Atman C, Fleming L, Miller R et al. Ac-
that is gathered, so there is need to find what understudies ademic pathways study: Processes and realities. In Proceed-
saying in the tweets. Along these lines, it first does an induc- ings of the American Society for Engineering Education An-
tive substance investigation on designing Problems dataset. nual Conference and Exposition. 2008.
Inductive substance investigation is one very much enjoyed
subjective research system for physically assessing content 4. R Ferguson. The state of learning analytics in 2012: A Re-
substance. view and future challenges. Knowledge Media Institute. Tech-
nical Report. KMI.2012.
Cite this article: Pridhivi Raj Kandula. Data Mining of Social Media Statistical Techniques for Analyzing Student’s Problems in Universities.
EMS Eng Sci j 2018, 2(1):009.
6/7
5. R Baker, K Yacef. The state of educational data mining in
2009: A review and future visions. Journal of Educational
Data Mining. 2009, 1(1): 3–17.

6. Tsoumakas G, Katakis I, Vlahavas I. Mining Multi-label data,


Data mining and knowledge discovery handbook. 2010,
667-685.

7. Dipak V Patil, Bichkar RS. Issues in Optimization of Deci-


sion Tree Learning: A Survey. International Journal of Applied
Information Systems. 2012, 3(5): 13-29.

8. AK Santra, S Jayasudha. Classification of Web Log Data to


Identify Interested Users. Using Naïve Bayesian Classification
in International Journal of Computer Science. 2012, 9(1):
381-387.

9. Manning CD, Raghavan P, Schtze H. Introduction to Infor-


mation Retrieval. New York, NY, USA: Cambridge University
Press, 2008.

10. Siemens G, Long P. Penetrating the fog: Analytics in learn-


ing and education. Educause Review. 2011, 46(5): 30–32.

Cite this article: Pridhivi Raj Kandula. Data Mining of Social Media Statistical Techniques for Analyzing Student’s Problems in Universities.
EMS Eng Sci j 2018, 2(1):009.
7/7

Potrebbero piacerti anche