Review of the Implications of Uploading Unverified Dataset in A Data Banking Site (Case Study of Kaggle).pdf

www.idosr.org Ezeet al
29
©IDOSR PUBLICATIONS
International Digital Organization for Scientific Research ISSN: 2550-7931
IDOSR JOURNAL OF APPLIED SCIENCES 7(1) 29-40, 2022.
Review of the Implications of Uploading Unverified Dataset in A Data
Banking Site (Case Study of Kaggle)
Val Hyginus U. Eze1,2
,Martin C. Eze1
, Chibuzo C. Ogbonna1
, Valentine S.
Enyi1
,Samuel A. Ugwu1
, Chidinma E. Eze3
1
Department of Electronic Engineering,University of Nigeria, Nsukka.
2
Department of Computer Science, Faculty of Science, University of Yaoundé 1,
Cameroon.
3
Department of Educational Foundation and Planning, University of Calabar, Calabar,
Nigeria.
ABSTRACT
This review paper comprehensively detailed the methodologies involved in data
analysis and theevaluation steps. It showed that steps and phases are the two main
methodological parameters to be considered during data assessment for data of high
qualities to be obtained.It is reviewed from this research that poor data quality is always
caused by incompleteness, inconsistency, integrity and time-related dimensions and the
four major factors that causes error in a dataset are duplication, commutative entries,
incorrect values and black entries which always leads to catastrophe. This paper also
reviewed the types of datasets, adopted techniques to ensure good data quality, types
of data measurement and its classifications.Furthermore, the Kaggle site was used as a
case study to show the trend of data growth and its consequences to the world and the
data bankers. It is then deduced that low data quality which is caused as a result of
errors during primary data mining and entries leads to wrong results which bring about
the wrong conclusions. It was advised that critical data quality measures should be
adopted by the data bankers such as Kaggle before uploading the data into their site to
avoid catastrophe and harm to humans.Finally, the outlined solutions as reviewed in
this paper will serve as a guide to data bankers and miners to obtain data of high
quality, fit for use and devoid of a defect.
Keywords: Accuracy, Data Bank,Data Quality, Dataset, Defect, fit for use, Kaggle
INTRODUCTION
The rapid growth of big data attracted
the attention of world researchers
towards accessing the quality of the
data to be used by an organisation
inorder to make it fit for a particular
purpose[1]. However, with the upswing
of technologies such as cloud
computing, the internetand social
media, the amount of generated data is
increasing exponentially[2]. The
enormous amount of data available in
many forms and types forces
organizations to come up with
innovative ideas to fine-tune data in-
order to maintain quality[3]. From
research, it was found out that data are
classified based on their structures.
Data with less quality are mainly
obtained in Unstructured data from
multiple sources which makes data
quality management complex. Some
causes of poor data
qualityarecarelessness in data entry,
unverified external web data sources,
system errors and poor data migration
processes. The increase in the amount
of data being used by organizations,
stored in data banks and mined for
competitive useover the last decades led
to a research shift from data mining to
data quality maintenance and integrity
in order to make it fit for use [4][5]. The
challenge forquality and quantityis
increasing daily as researchers are
always in need of it for a betteroutput.
High-quality data is a prerequisite for
worldwide research harmonization,
global spend analysis and integrated
service management in compliance with
regulatory and legal requirements [6][7].
The level of poor-quality data being
uploaded in some data mining sites has
awakened the spirit of researchersin
refining the quality of the data for better
results. However, the use of poor-quality
dataisso catastropheand leads to
inaccurate results, inefficiency, risk

30
mitigation and wrong decisions
making[8].
Data Quality
Dataset involves the collection of
related data items that may be accessed
individually or in combination or
managed as a whole entity. A
databasecan be structured to contain
various information about a particular
category of relevant to the data
researcher. A database can also be
considered asa datasetas some entities
within it are related to a particular type
of information[9][10][11][12]. Data
quality is the high-quality information
that is fit for use. They are said to be fit
for use when they are free of defects
and possess the features needed to
guide the user in making the right
decision.It can also be expressed as the
discrepancy between the way data were
viewed when presented as a piece of
information in a system and
itsappearance in real-life
scenarios.Hence, data quality can be
generally expressed as the set of data
that is fit for use and free from
defects.For data to be certified fit for
use and free from defectsit must be
characterized and examined under the
data quality dimension.
Factors that Determinesthe Data Quality
The key major factors of data quality
dimensions that determine the fitness of
data to be used are; data accuracy, data
completeness, data consistency and
time-related dimensions [13].
Data Accuracy
Accuracy is one of the data quality
dimensionsthat is widely used to
ascertain the fitness of data for use.
Accuracy is the closeness of a data value
and its real-lifevalue (real-world
object)which the data aims to
represent.Data accuracy can be
classified as either syntactic accuracy or
semantic accuracy.
 Syntactic accuracy is the
closeness of a data value to the
elements of the corresponding
definition domain. In syntactic
accuracy, a data value is not
compared to the real-world
object that it aims to represent
rather it focuses on checking the
correspondence of a data value
with respect to any value in the
domain that defines the data
value.
 Semantic accuracy is the
closeness of a data value and the
real-world object which it aims to
represent. In order to be able to
measure semantic accuracy, the
true value of the real-world
object must be known for
comparison [14].
Data Completeness
Completeness is the extent to which
data are of sufficient breadth, depth and
scope for the task that is meant to solve.
It isclassified into schema
completeness, population completeness
and column completeness [15].
 Schema completeness is the
degree to which concepts and
their properties are not missing
from a data schema.
 Population completeness
evaluates missing values
withrespect to a reference
population.
 Column completeness is defined
as a measure of the missing
values for a specific property or
column in a dataset.
When it is important to know why a
value is missing and also to know when
a value is represented with NAN or NA.
In measuring the completeness of a
table, it is important to note the
difference between null values and
incompleteness, NAN and NA.
A value is said to null value when it
possesses either of these characteristics:
 The value is not existing (does
not contribute to
incompleteness)
 The value is existing but not
known (contributes to
incompleteness)
 It is not known whether the value
exists (may or may not contribute
to completeness)

31
From this, it can be concluded that
incompleteness (missing value) can be
detected but incompleteness cannot be
certainly solved accurately unless the
data miner is present to ascertain which
type of error existed in that dataset or
the original datasheet is available.
Data Consistency
Data consistency involves semantic rule
violation which is expressed as the
integrityconstraints of dataset
properties that must be satisfied for
effective performance.There are two
fundamental categories of integrity
constraints known as intra and inter-
relation constraints.
 Intra-relation constraints define a
range of admissible values for an
attribute where such contrary
values do not exist in that
contest. For example, a negative
age in a database represents a
person.
 Inter-relation constraints involve
attributes from other relational
databases.For example,an
individual of different ages is
identified in two databases.
Time-Related Dimensions of Data
Data quality can be characterized by the
ability of the dataset to maintain its
quality after some years of
metamorphosis and also how current it
is up to date. Most research recognizes
three closely related time dimensions
such as currency, volatility and
timeliness. The three time-related
dimensions and their relationship based
on their characteristics are expressed as
tohow often the database was updated
or the time between receiving a data
unit and the delivery of the data unit to
a customer.
The relationship between the three time-
related dimensions is mathematically
expressed as in equation (1).
Timeless = max (1)
Data Dimension Measurements
Data Dimension Measurement (DDM) is
one of the important criteria to be
considered in validating the quality of
data. To design the right metrics to be
adopted in measuring data quality is
one of the most challenging tasks of
data quality assessment as it
shouldidentify all errorswithout
reflecting or repeating the same errors
multiple times [16]. The simplest
metrics used in obtaining the value of
objective measures which is expressed
as the ratio of the error line to the total
lines of the dataset as shown in
equation (2).
Ratio = (2)
However, the calculation of such ratios
is only possible when there are clear
rules on when an outcome is desirable
or undesirable which forms the basic
foundation for a good outcome[15].It is
imperative to note that most methods
provide only objective measures for
assessing data quality dimensions
without recognizing the importance of
the distinction between subjective and
objective measures and the comparison
between themwhich forms the basic
input for the identification of data
quality problems. Table 1 showed the
differences between objective and
subjective measures.
Table 1: Objective Versus Subjective Data Quality Measures
Feature Subjective Objective
Measurement tool Software Survey
Measuring target Datum Representational information
Measuring standard Rules, patterns User satisfaction
Process Automated User involved
Results Single Multiple
Data storage Databases Business contexts

32
Types of Datasets
The supreme objective of any Data
Quality analyst is to analyse and
evaluate the quality of data in
useaccurately, to ensure that it is fit for
use and devoid of a defect. The concept
of data itself involves a digital
representation of real-world objects that
can be later processed and manipulated
by software procedures through a
network. Data analysis is the process of
combining extracted information and
data mining techniques to achieve
higher quality. Information extraction is
a process of populating a database
fromunstructured or loosely structured
text. Data mining techniques are being
applied to discover patterns that can be
later assessed for better results.
In the field of information systems, data
are being grouped as follow:
 Structured
 Unstructured
 Semi-structured.
In the field of statistics, data are
alsobeing grouped as:
 Categorical
 Numerical
 Ordinal.
Categorical or Structured Data
This is also called structured data in
information science which is expressed
asa group of item(s)with its simple
attributes defined within a domain. A
Domain is the type of value that can be
assigned to an attribute. Structured data
is categorically referred to as
corresponding programming language
data types such as Integer (whole
numbers), Boolean (true/false), String
(sequence of characters). Moreover, in
statistics it is called Categorical
datawhich represents some data
characteristics such as a person‟s sex,
marital status, age etc. Categorical data
can also use numerical values to
represent strings such as 1to represent
true and 0 to represent false.The
assigned values have no analytical
values as they cannot be added or
subtracted together [17][18].
Numerical or Unstructured data
Unstructured data are ungrouped and
non-tabulated sets of data presented in
a well acceptable language such as news
articles, web pages, blogs, e-mails, etc.
Unstructured data could not only be text
but also images, audio and video clips.
A huge amount of unstructured data is
easily obtained and also available on the
internet nowadays in which there is a
need to analyse and process such data
for future use. The accuracy of
structured data is dependent on the
final user as the accuracy of structured
data will be higher when a final
consumer is a machine than when it is
human. Information extraction consists
of five substantialsub-tasks as
segmentation, classification,
association, normalization, and
reduplication and in statistics,this is
calledNumerical data. These data have
the meaning as a measurement such as
height, weight, IQ, blood pressure.
Statisticians also call numerical
data quantitative data and it is also
divided into discrete and
continuousnumerical data.
a. Discrete data represent items that
can be counted and also take a possible
value that can be listed out. The list of
possible values may be fixed (finite).
b. Continuous data represent
measurements whose possible values
cannot be counted and can only be
described using intervals on the real
number line.For example, the exact
amount of gas purchased at the pump
for cars with 10 litres tanks would be
continuous data from 0 litres to 10 litres
which can be represented by the interval
[0, 20][19].
Ordinal or Semi-Structured data
This is called semi-structured data in
information science and it is expressed
as flexiblystructured data with a semi-
similar characteristic in comparison
with traditional structured data. It is
expressed to be flexible due to its
inability to adopt the formal structure of
data models associated with relational
databases or data tables. Ordinal data
statistically mixes both numerical and
categorical data together. The data fall
into categories, but the numbers placed
on the categories have meaning. For
example, grading students in an exam
on a scale from lowest (0) to highest (70)
scores gives ordinal data. Ordinal data
are often treated as categorical, where
the groups are ordered when graphs and
charts are made. For example, if you

33
survey 100 people and ask them to rate
an exam on a scale from 0-70, taking the
average of the 100 responses will have
meaningbutvaries with categorical
data[18]
Types of Data Measurement Scale
There are two major data measurement
scales used in measuring the accuracy
of calibrated data measuring
instruments. The two most popular data
measurement scales are ratio and
interval scales.
Interval scales are numerical scale that
is used to indicate the space or
difference between two
variables/values. It is so important as
the realm of statistical analysis on the
datasets are open for measurement and
analysis. For example,a central tendency
can be used to measure mode, mean,
median and standard deviation. The
drawback of this is that it does not have
a true zero.
Ratio scale is an ultimate nirvana when
it comes to data measurement scales
because it gives detailed information
about the order and exact unit values
and also has absolute zero which gives
it an edge for measuring a very wide
range of values of descriptive and
inferential statistics [20].
Classification of Data
Data classification involves the methods
and patterns at which data were
collected and organized for use. Data
can be broadly classified based on
where and how they are gotten(sourced)
which are; primary and secondary data
sources.
1.Primary Data:A directly sourced data
that has not been altered, refined or
published isknown as primary
data.Primary data could be sourced
from experiments, one on one
interviews, questionnaires etc. An
example of such isdata obtained from
renewable energy research on the effect
of dust in the solar photovoltaic panel
that was performed practically and data
collected by the researcher.
2. Secondary data:Secondhand
information or refined or altered data
collected and recorded for future use is
known as secondary data. It can also be
a piece of firsthand information that
was recorded by the researcher for
future use by other researchers. These
secondary data sources are found in
some software applications/sites such
as Jamia, Alibaba, jigjig andKaggle. The
owner of such data might have refined it
to suit his purpose and finally lunch it
for researchersto consult for
information with respect to their
research area.
The problem of secondary data which is
refining and fine turning of datasets has
caused more harm than good in the
research field. The refined data causes
problems as the refiner will not
categorically state what has been refined
so that the user will be aware that some
of the inbuilt microns of that datahave
been altered. This data might be used
for research purposes even without
rugged analysis since the secondary
user lacks the knowledge of where these
data was sourced from and the
authenticity of it but relies on the
description of the publisher. In some
cases, some values/data uploaded can
be autogenerated and were prone to
errors and manipulations.This has made
secondary data highly unreliable yet
doesn‟t discredit the use of secondary
data as it has merits of easier access,
aids in making research faster and it is
generally cheaper thana primary source
of data.
Data Quality Assessment
Configurations
There have been difficulties in
ascertaining the best data quality
configurations to be adopted during the
data processing.Some researchers
proposed a dynamic configurational
data quality assessment type whose
input configuration is generic data
containingcritical activities. This generic
assessment technique was formulated
by harnessing and grouping activities
from selected data quality assessment
methodologies.The order and
dependencies between activities were
defined based on the input and the
output of the specific activities.Generic
data quality process and configuration
methods were defined to have the
following characteristics:
1. Determining the aim of the
assessment and requirements related to
the data quality assessment. The
assessment status of the accessed
processes is so important to be stated

34
by the data quality assessors to guide
users on how, when and where a
particular method should be usedfor
fitness.
2.Select the activities from the generic
process model that contribute to the
assessment aim and requirements
3.Configure and arrange activities in a
sensible order and include activity
dependencies.
Figure 1 showed the generic data quality
assessment processes adopted by
different data quality assessors to
ensure that the data in use is fit for the
purpose.
Comparative Analysis of Data Quality Assessment Methodologies
This section of the research paper
reviewed a comparative analysis of data
quality assessmenttechniquesthat
existed as of the time of this
compilation with the aim of identifying
critical activities in dataquality.The
thirteen existingdataquality assessment
and improvement methods based on
steps, strategies, techniques,
dimensions and types of information are
as shown below:
 The methodological phases and
steps
 The strategies and techniques
 The data quality dimensions and
metrics
 The types of data in use
 The types of information systems
in use [21].
The methodology comparison measured
on phases and steps is of utmost
interest ina comprehensive review of the
existing methods as phases and
stepswere so important in ascertaining
data quality.Table 2 showed the most
effective and organised steps
recognized in data quality assessment
processes. The table further analysed
data based on sources and related
problems associated with each method
used indata sourcing. There are set up
rules used by schemas to examine and
interview clients in order to improve
performance and achieve a complete
understanding of data and its related
architectural management. The
following steps are incorporated in
order to obtain effective and reliable
data. Important steps to be adopted for
Figure 1: A generic data quality assessment process

35
better quality data output to be
achieved are as outlined below:
1. Data quality requirement
analysis:This is a technique where the
opinion of data users and
administrators are surveyed and
sampled to identify data quality issues
and set new quality targets.
2. Identification of critical areas:This
involves the identifications of the
relevant areas and the sequence of data
flow in a database for easy location and
access.
3.Process modelling:This is the process
of validating and updating produced
data for fitness
4. Measurement of quality:This
involves the use of different data quality
dimensions, the problems associated
with each dimension quality and the
definition of the dimensional metrics to be
used.
Table 2 Comparison of Methodologies and their Assessment steps
Step/meth
acronym
Data
analysis
DQ
requirement
analysis
Identification
of critical
areas
Process
modelling
Measurement
of quality
Extensible to
other
dimensionsand
matrices
TDQM + + + + Fixed
DWQ + + + + Open
TIQM + + + + + Fixed
AIMQ + + + Fixed
CIHI + + Fixed
DQA + + + Open
IQM + + Open
ISTAT + + Fixed
AMEQ + + + + Open
COLDQ + + + + + Fixed
DaQuinCIS + + + + Open
QAFD + + + + Fixed
CDQ + + + + + Open
From this comprehensive review of
methodologies, it was observed that no
method is the best or superior to others
for data quality assessment but there
are steps and phases to combine or
follow to obtain a very high output.
Causes of Error in Dataset
The causes of error in a dataset that
always leads to low data quality are as
detailed below:
Commutative entries: This is a common
issue where values are wrongly placed
in an attribute that was not meant for it.
It is a fake entry that might be as a
result of omission or sourcing data from
a means that are not fit for its purpose
and to detect such error is very difficult
especially in cumbersome data.
Incorrect values: This involves
entering values either out of range or
incorrectly which can cause grave errors
in calculation. For instance, using
numeric values for numeric attributes,
„O‟ instead of 0 in digits [22] with the
human eye, it will be difficult to detect.
Blank entries: This is mostly found in
structured CSV/excel and Relational
databases. For example, age in some
cases, can‟t be a primary key, and as
such blank space can be permitted. This
is not appropriate especially when the
researcher or user needs accurate age
for research work. Every missing or
blank space in a dataset is necessary to
be filled with a value that will not
change/negate the quality and
characteristics of that dataset.
Duplication: This can be seen in some
data entries where a row or column
entry was duplicated once or more. It
mainly occurs where data are
automatically generated and leads to
false outcomes.
Causes of Low Data Quality
Inconsistency:This involves semantic
rule violation which is expressed as the

36
integrity constraints of a dataset
property that must be satisfied for
effective performance.
Incompleteness:This is the extent to
which data are insufficient in breadth,
depth and scope for the task that is
meant to solve.
Inaccuracy:The rate of variance of the
closeness of a data value and its real-life
value (real-world object) which the data
aims to represent.
Duplication: This occurs when the same
data with the same characteristics are
being enteredonce or more in a row or
column.
Blank Entries: This is a process where a
column or row is total empty where
there is a certainty of having a value.
Incorrect values: This involves using a
wrong symbol that resembles in place of
anothercorrect one.
Commutative Entries: This is a common
issue where values are wrongly placed
in an attribute that was not meant for it.
Furthermore, many data bankers like
Kaggle, Jumia, Konga etc should be
cognizant of these aforementioned
causes and characteristics of poor data
quality and ensure that data uploaded in
their sites should be devoid of such
casualty.
Kaggle Data Site
Kaggle is an online data science
community where passionate data
science researcherslearn and exchange
ideas [23]. Kaggle webpage can be
accessed via https://www.kaggle.com/.
Most researchers more especially data
scientists findthe Kaggle site very useful
for data sourcing even though it has it's
own prone and cons[24][25]. Since it is a
community, it encourages sharing
datasets publicly among users. Datasets
can also be private which enables the
user to use them privately. It can be
shared personallywith another user
alongside its security key. There are
stipulated guidelines on how to create
datasets, rate datasets, ask for
collaboration and in general how to use
these datasets [26]. Datasets in Kaggle
appears on different data formats,
depending on which format the users
deem fit to use. These formats range
from CSV, JSON, xlsx, Archives to
BigQuery, with CSV as the most available
and used format. The challenge of being
unable to check and ascertain the degree
of the data correctness, completeness,
integrity and accuracy together with the
quality of data to be uploaded in the
Kaggle site even though the data owners
have a descriptive space where to write
a brief detail of the uploaded data has
caused the major concern. This is so
because many of the data uploaded in
the site were refined and altered by the
data miner without stating exactly what
was altered and why it was refined in
the description space
Kaggle Data Qualities
Kaggle is a worldwide site where huge
and bulk data were obtained for
scientific and machine learning research
purposes. For this site to attract more
customers and researchers to be visiting
and always confide on the uploaded
data in the site certain quality assurance
will be needed by the site users such as
the quality of data to be uploaded and
the sources of the data. However,
because of the popularity of the Kaggle
site, several data are being uploaded
without proper debugging and cleaning
which degrades the quality of data on
the site. This aforementioned drawback
led to ongoing data quality research on
Kaggle and other data banking sites.
Data quality can generally be expressed
as a data‟s fitness for use and also its
ability to meet up the purpose set by the
data user. This definition simply
showed that the quality of data is highly
dependent on the context of the data
userin synergy with the customer needs,
ability to use and ability to access data
at the right time and at the right
location. However,the data quality
assessment and improvement
processare not limited to the primary
data miners only but also to the data
users and other data stakeholders that
are involved during data entry, data
processing and data analysis [27]. Unlike
the conventional method of data
division and analysis where synergies
and coherency among the data
managers were not inculcated and the
bridge between them seriously affects
the quality of uploaded data. In Kaggle
data quality is divided into four
subgroups known as data quality for a
website, data quality for decision
support, data quality assessment and

37
other data quality applications such as
medical data quality and software
development processes for engineers.
Furthermore, the study has proofed that
data quality assurance is the major
problem ofthe Kaggle site. Many
researchers have studied the data‟s in
Kaggle and discovered that at first when
data were not as huge as in Tera Bytes
the challenges was limited to
completeness and accuracy.However, as
data began to grow in size and quantity
the research and challenges which is
peculiar to the field of study and
research area shifted from two to five in
the field of data management and
otherswhichare completeness,
correctness, consistency, plausibility,
and currency. Therefore, accuracy,
completeness and consistency mostly
affect the capacity of data to support
research conclusions and therefore form
the most important area of research in
the data mining sector[28].
Trends and Importance of Data
Quality Assurance in Kaggle
There have been numerous attempts to
summarize the research on data quality
trends in Kagglebased on the best
methodology to adopt in examining the
quality of data but all to no avail. Then,
after the industrial revolution, the
amount of information dominated by
characters doubled every ten years.
After 1970, the amount of information
doubled every three years. Today, the
global amount of information can be
doubled every two years. In 2011, the
amount of global data created and
copied reached 1.8 ZB. It is difficult to
collect, clean, integrate, and finally
obtain the necessary high-quality data
within a reasonable time frame because
of the proportion of the high
unstructured data in a big dataset which
takes a lot of time to be transformedinto
a structured type of dataset and further
processed. This is a great challenge to
the existing techniques of data quality
processing and management [29].
The first seminar data quality research
topic was introduced by [30]in the year
1995. Sincethen, comprehensive
research was going on to summarize,
classify, and develop frameworks for
Data Quality research.[31][32][33]. In
[30] a framework of data quality from a
comprehensive analysis of publications
from 1994-1995 was proposed. The
authors compared data and data quality
to be a manufactured physical product
and its quality and relates it to
managing data quality from established
concepts to managing the quality of
physical products. Moreover, research
was still ongoing as the quantity of data
in Kaggle kept on growing every minute
of the day which lead to a summary of
data quality research using articles
published between 1995 and 2005[31].
However, from reviewed Kaggle data and
others, relationships werederived based
on the judgment and intuition of the
researchers‟ present conceptual
assessment of data quality and its
management forexcellent results.
According to[34], data quality research
can be classified categorically based on
assessment, management, and
contextual aspects of data quality. Data
quality as a novel framework that
combined the factors of fitness for
useas defined in [35] and with the
incorporation of management elements
as defined in[30] is comprehensively
reviewed in [32]. Furthermore, many
topics on methods to categorize data
quality research and to develop a
framework that allows researchers to
characterize their research was also
reviewed on[33]and it showed different
taxonomy and methods at which data
can be examined for quality assurance.
In summary, as stated earlier, the data
quality research in Kaggle has grown to
a critical juncture that has attracted the
interest of researchers and the world at
large. From 1907 to 2010, data quality
was considered somewhat anecdotal and
esoteric. But as of 2021, it is considered
valuable and relevant because of the
importance of datato researchers, data
scientists and the world at large. Due to
the tremendous upload of data and
bigdata in Kaggle, the research area has
witnessed extraordinary growth in
thelast five years and as a result of that
Kaggle site serves as one of the best
data banks for researchers.Many data
bankers like Kaggle, Jumia, jigjig etc will
benefit so much from this review paper
as it will help them in configuring data
to be uploaded to their site based on the

38
reviewed methodologies for high output
to be achieved.
CONCLUSION
This review showed the different
methods and techniques that can be
adopted by researchers to ensure that
data of high quality is obtained. It also
detailed the types of data, data quality
assurance, causes of poor data quality
and the consequences. It showed the
different domains at which data exists
and also extensively detailed the types
and classes of datasets that exist and
their importance to data users. Kaggle
data site was used as a case study in
this review paper and it detailly
reviewedthe different errors
encountered by Kaggle and highlighted
the causes of it. This reviewwill serve as
a guide to data scientists, primary data
miners and data bankers on the best
ways to handle their data to be devoid
of defect, error and also to make it fit
for use at all times. Finally, this paper
will assist data bankers in checking for
quality of data, data quality assurance
and the authenticity of the data from
the primary miners before uploading it
into their sites to avoid misleading
researchers or data users.
REFERENCES
[1] “Abbasi, A., Sarker, S., & Chiang,
R. H. L. (2016). Big Data Research
in Information Systems: Toward
an Inclusive Research Agenda.
Journal of the Association for
Information Systems, 17(2), 1-33.
[2] “Cai, L., & Zhu, Y. (2015). The
Challenges of Data Quality and
Data Quality Assessment in the
Big Data Era. Data Science Journal,
14(0), 2.
[3] “Albala, M. (2011). Making Sense
of Big Data in the Petabyte Age.
Cognizant 20-20 Insights
Executive Summary,” p. 2011,
2011.
[4] “Cappiello, C., Francalanci, C., &
Pernici, B. (2004). Data quality
assessment from the user‟s
perspective. Proceedings of the
2004 International Workshop on
Information Quality in
Information Systems, 68–73.
[5] “Hevner, A. R., & Chatterjee, S.
(2010). Introduction to Design
Science Research. In Design
research in information systems:
Theory and practice (Vol. 28, pp.
1–8).
[6] “Kagermann, H., Österle, H., &
Jordan, J. M. (2011). IT-Driven
Business Models: Global Case
Studies in Tranformations. John
Wiley & Sons.”
[7] “Shankaranarayan, G., Ziad, M., &
Wang, R. Y. (2003). Managing Data
Quality in Dynamic Decision
Environments: An information
product approach. Journal of
Database Management, 14(4), 14–
32.
[8] “Friedman, T., & Smith, M. (2011).
Measuring the Business Value of
Data Quality. Gartner,
(G00218962).”
[9] TechTarget, “data set,”
www.whatis.com,
https://whatis.techtarget.com/def
inition/data-set2020
[10]
https://blog.ldodds.com/2013/0
2/09/what-is-a-dataset/Ldodds,
“Definitions of Dataset6,”
www.blog.ldodds.com,2013
[11] byjus.com,” 2019.
https://byjus.com/maths/data-
sets/.
[12] “J Uran, J. M., & Godfrey, A. B.
(1998). Juran‟s Quality Handbook.
McGraw-Hill. McGraw Hill.
[13] “Scannapieco, M., & Catarci, T.
(2002). Data Quality under the
Computer Science perspective.
Computer Engineering, 2(2), 1–
12,” vol. 2, p. 2002, 2002.
[14] “Batini, C., & Scannapieco, M.
(2006). Data Quality Concepts,
Methodologies and Techniques.”
[15] “Pipino, L. L., Lee, Y. W., & Wang,
R. Y. (2002). Data quality
assessment. Communications of
the ACM, 45(4),vol. 45, p. 506010,
2002.
[16] “del Pilar Angeles, M., & García-
Ugalde, F. J. (2009). A Data Quality
Practical Approach. International
Journal on Advances in Software,
2(3).,” vol. 2, p. 2009, 2009.
[17] T. Kupta, “Types of Data Sets in
Data Science, Data Mining &

39
Machine Learning,” www. @
towardsdatascience.com, 2019.
https://towardsdatascience.com/t
ypes-of-data-sets-in-data-science-
data-mining-machine-learning-
eb47c80af7a.
[18] K. Nandi, “Explore Your Data:
Cases, Variables, Types of
Variables,” www.web@
makemeanalyst.com, 2018.
http://makemeanalyst.com/basic-
statistics-explore-your-data-cases-
variables-types-of-variables/.
[19] N. Donges, “Data Types in
Statistics,”
www.towardsdatascience.com,
2018.
https://towardsdatascience.com/
data-types-in-statistics-
347e152e8bee.
[20] L. L, “Types of Data &
Measurement Scales: Nominal,
Ordinal, Interval and Ratio,”
www.mymarketresearchmethods.c
om, 2020.
[21] “Batini, C., Cappiello, C.,
Francalanci, C., & Maurino, A.
(2009). Methodologies for data
quality assessment and
improvement. ACM Computing
Surveys, 41(3), 1–52. vol. 41, p.
1541883, 2009.
[22] “P. Bhandari, „Scribbr-What is a
ratio scale of measurement,‟
Scribbr [Online]. Available:
https://www.scribbr.com/statistic
s/ratio-data/. [Accessed 28
October 2020].”
[23] “Wikipedia, „Kaggle,‟ Wikipedia,
[Online]. Available:
https://en.wikipedia.org/wiki/Kag
gle. [Accessed 9 October 2020].,”
no. October 2020.
[24] “„Quora - How appropriate it is to
publish research-based on Kaggle
competitions data on ArXiv,‟
Quora, 25 January 2016. [Online].
Available: https://qr.ae/pNkBBI.
[Accessed 30 October 2020].”
[25] “V. Nikulin, „An On the Method for
Data Streams Aggregation to
Predict Shoppers Loyalty,‟ IJCNN,
pp. 1454-1461, 12 July 2015.,” no.
July, p. 2015, 2015.
[26] “Kaggle, „Kaggle Datasets,‟ Kaggle,
[Online]. Available:
https://www.kaggle.com/docs/da
tasets. [Accessed 2
November2020].”
[27] “A Review of Data Quality
Research in Achieving,” Journal of
Theoretical and Applied
Information Technology, vol. 95,
no. 12, pp. 1–12, 2017.
[28] N. Z. Meredith, W. E. Hammond, B.
G. Beverly, and M. G. Kahn,
“Assessing Data Quality for
Healthcare Systems Data Used in
Clinical Research,” Health Care
System Research Collaboratory,
vol. 1, pp. 1–26, 2014.
[29] L. Cai and Y. Zhu, “The Challenges
of Data Quality and Data Quality
Assessment in the Big Data Era,”
Data Science Journal, vol. 12, no.
2, pp. 1–10, 2020.
[30] “Wang, R. Y., Storey, V. C., and
Firth, P. 1995. „A Framework for
Analysis of Data Quality
Research,‟ IEEE Transactions on
Knowledge and Data Engineering
(7), pp. 623-640.,” no. 7, p. 1995,
1995.
[31] “Lima, L., Maçada, G., and Vargas,
L.M. 2006. „Research into
information quality: A study of
the state-of-the-art in IQ and its
consolidation,‟ in Proceedings of
the International Conference on
Information Quality, Cambridge,
MA.,” p. 2006, 2006.
[32] “Neely, M. P. and Cook, J. 2008. „A
Framework for Classification of
the Data and Information Quality
Literature and Preliminary Results
(1996-2007),‟ in Americas
Conference on Information
Systems (AMCIS), Toronto, CA,”
2008.
[33] “Madnick, S., Wang, R. Y., and Lee,
Y. W. 2009. „Overview and
Framework for Data and
Information Quality Research,‟
ACM Journal of Information and
Data Quality, (1), pp. 1-22.,” no. 1,
2009.
[34] “Ge, M. and Helfert, M. 2007. „A
Review of Information Quality
Research,‟ in Proceedings of the
International Conference on
Information Quality, Cambridge,
MA.,” p. 2007, 2007.
[35] “Juran, J. M. and Godfrey, A. B.
2000. Juran‟s Quality Handbook,

40
McGraw Hill International
Editions: Industrial Engineering
Series, 5th Edition, 2000.

Review of the Implications of Uploading Unverified Dataset in A Data Banking Site (Case Study of Kaggle).pdf

More Related Content

Similar to Review of the Implications of Uploading Unverified Dataset in A Data Banking Site (Case Study of Kaggle).pdf (20)

More from ssuser793b4e (20)

Recently uploaded (20)

Review of the Implications of Uploading Unverified Dataset in A Data Banking Site (Case Study of Kaggle).pdf