Volume 6, Issue 10, October – 2021 International Journal of Innovative Science and Research Technology

ISSN No:-2456-2165

A Framework for Improving Key Performance

Indicators using Business Intelligence Techniques
Abdel Nasser H. Zaied Dr. Riham Hagag Assem Khalaf Ahmed
Prof., Professor of Systems Prof., Information systems Business Information Systems
Engineering (Information Systems), Faculty of commerce department
Former Dean, College of Computers Helwan University Faculty of commerce and business
and Informatics, Cairo, Egypt administration
Zagazig University, Egypt. Helwan University Cairo, Egypt
Cairo, Egypt

Abstract:- The study aims to identify the most important II. BUSINESS INTELLIGENCE CONCEPTS
factors affecting the main performance indicators in
business intelligence inside Business Organizations, by A. Definition
classifying them according to the degree of impact on Business intelligence (BI) alludes to the procedural and
them. technical foundation that gathers, stores and analyzes the data
delivered by an organization's activities. Business insight is a
The Data mining methods help to extracts the wide term that incorporates information mining, measure
hidden relation between the data in order to analysis, and examination, execution benchmarking, engaging
then predict information from large databases. One of investigation, etc. Business insight is intended to take in all
the most important tools used in data mining is called the information being created by a business and present
neural networks. It works through learning and simple to process execution measures and patterns that will
prediction. The higher the input values, the greater the illuminate the executive's choice [6].
learning process because it depends on training.
B. Tasks
Keywords:- Business Intelligent, Data Mining, Knowledge The expected advantages We find that all elements and
Discovery, Key Performance Indicators. tools of business intelligence aim to provide and improve
decision-making, simplify and speed up internal processes
I. INTRODUCTION and add an increase operational productivity, driving new
incomes and gaining competitive advantage over business
Dashboards are regularly made out of data display rivals.
technology, including Key Performance Indicators (KPIs)
which assume an essential part in rapidly giving precise data BI frameworks can likewise assist organizations with
by looking at current execution against an objective needed to distinguishing market patterns and spot business issues that
fulfil business targets. In any case, KPIs are not in every case should be tended to [12].
notable and in some cases Hard to find a proper KPIs to
connect with every It aims to work [11]. Business Intelligence frameworks give authentic,
current, and predictive perspectives on business activities,
Business Intelligence frameworks Grants verifiability frequently utilizing information that has been accumulated
and current views of any business activity, we find a lot of into data warehouse or data mart and sometimes working
information used and collected as a whole in the data from operational data [5]. Programming components uphold
warehouse and work intermittently with any information announcing, intuitive "cut up" turn table investigations,
during the operation process. The components of the perception, and factual information mining. Applications
Software elements support the preparation of reports tackle deals, creation, monetary, and numerous different
Interactive pivot table analytics and also mining for wellsprings of business information for purposes that
measurable information [12]. incorporate business execution the executives. Data is
regularly accumulated about different organizations in a
It provides processing applications, whether sales, similar industry which is known as benchmarking [14]. BI
production, finance, and various other sources of business information the stored data warehouse includes historical
data, and its purpose includes business performance information, like the new and modern information that is
management. collected when it is created from the source of the
frameworks, to enable the tools used in Business Intelligence
Data is accumulated regularly about different to provide support for all major, strategic and dynamic
organizations in a similar industry and known as performance processes.
measurement.

IJISRT21OCT210 www.ijisrt.com 295

Volume 6, Issue 10, October – 2021 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
III. DATA MINING CONCEPTS is called the information layer, and the last layer is called the
return layer and Between the first and the last layer there is
A. Definition only one layer.
Data mining to enhance decision-making through a
range of computer technologies and applications, this will be Data flows forward relative to the feed forward in one
by applying prediction and description methods to a large direction from the first layer to the last layer. This type of
amount of data to increase decision-making ability, increase network works by what is called a backscattering learning
accuracy, reduce analysis time, reduce poor performance algorithm. MLP are commonly is used configuration request,
indicators used, and increase the decision-making process affirmation, (MLP) can tackle issues which cannot be
through it because it is extracted from Through cached data distinguished directly [28].
neural networks that provide business intelligence assistance
in organizations [7]. 2) Neural Network (Probabilistic):
Class problems Here we can use (PNN). If an entry is
B. Tasks submitted, the distances in the initial layer are calculated
A large number of tasks suitable for the use of records starting from the input vector and ending with the training
in general will be collecting, evaluating, forecasting, vectors and producing a direction indicating the direction
estimating, and clustering, with outlining. An assortment of from the training input to the center [29].
them will in general be best drawn nearer in the top-down
design just although, in general, an incremental In the second layer, the contributions of each class of
approximation is preferred, called discovery, which often inputs are combined so that its net output is directed to the
leads to cancellation of regression [10]. possibilities. And then on most of those possibilities, the
feature of switching outputs appears. You have a choice of
With respect to Classification, is the highest well- the second layer options that produce either one for a specific
known information mining undertaking and it comprises of category or zero for any other category.
analyzing the highlights of a recently introduced object to
appoint for a pre-selected group of categories [18]. Probabilistic Neural Network It is an architecture of
neural networks that is useful and has a well-defined basis
However order manages Partial results, with results of with little backpropagation, but it feeds the structure forward
continuing value, appreciation is treated. In fact, evaluation is to be similar to the posterior diffusion. PNN is supervised
used regularly to perform a classification task [24]. learning set of rules but consists of and in the hidden layer,
there are no weights [30].
Expectation manages the grouping of records as
indicated by some anticipated future conduct or assessed 3) Neural Network (Linear):
future worth. Both are close gathering and mart crate (LN) provides a benchmark by which we can examine
examination with as a target to decide Things that can be put the overall performance of a neural network. In other cases, a
together. Grouping targets sectioning a heterogeneous complex problem can be solved by means of a linear network,
populace into various many homogeneous groups and as happens in neural networks. It is used in the case of a small
subgroups are also not previously identified. number of training cases, and here you do not need a model
that is more complex.
The goal is to present data well, so the complex
operations that take place in databases are described and In the case of a specific input with vectors, the result is
explained [25]. an output directed to the target it corresponds to. Each vector
has an input, and we hope for the time from calculating an
C. Data Mining (Classification Techniques) output vector to the network. Each of them differs in error in
Here we find, a set of records in databases is assigned relation to the output vector and the input vector. In the event
to a sample class or tag for classification from a large set of that you want to find a value for each weight in the network
predefined class labels. For example, the organization will be to reduce all the error boxes to a minimum and to the lowest
eager to classifying each of them. All the options, as well as value. The problem is that linear networks have simple errors.
the yes or no option, belong to classification problems. The For most cases, we can compute a linear network
rating has at least more than one level, for example, "high," immediately, with the error being the lowest of the input
"medium," and "low." The central matter the number of vectors [30].
classes is limited. It is noted who can be there a verifiable
request we find the relationship inside the definition of each 4) Neural Network (Radial Basis Function):
layer, for example, "high," "medium," and "low." [27]. Neural Network (Radial Basis Function) it is based on
a supervised learning algorithm which is pre-fed. Usually it
Neural Network Proposed Techniques: consists through one hidden layer of a group of tools to
choose to activate its own function through some of the
1) Multilayer Perceptrons Neural Network features referred to called base features.
Multi-Layer perceptron (MLP) is a stock forward neural
A network consists of at least one layer of the input layer as
well as the output layer, The first layer is the main layer and

IJISRT21OCT210 www.ijisrt.com 296

Volume 6, Issue 10, October – 2021 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
This network has many advantages, it works in a faster IV. PROPOSED SOLUTION
way and is less prone to input problems and the reason is the
radiation behavior of each hidden unit. A. A Framework of improving key performance indicators
using Business Intelligence Techniques
The design of that network consists of three layers, but The first figure shows the real A Framework for
each part is separate from the other, which is a group of nodes improving key performance indicators in the research paper.
in the case of input and in the case of output, so the reaction That the mining Techniques implemented the improving key
of the network to each type of input layers or the second layer performance indicators through the use of neural networks
is hidden and its dimensions are redundant. and Comparison the Classification results of Neural
Networks, the Radial basis Function, Multi-Layer Perceptron,
It constitutes a class of different artificial neural Linear and Probabilistic Neural Network to choosing the best
networks, and its advantages over other different networks result of neural network in data classification, for Application
appear in terms of simple form and study in a faster manner. of best Key performance indicators (KPIs) by using Business
These types of networks depend heavily on an estimate of intelligence Techniques to enhances to effectiveness to
many of the parameters: weights, centers, and widths that link support and decision making.
through the neural network [7].

Fig. 1. A Framework for improving key performance indicators using Business Intelligence Techniques

B. A Framework stages: That way is followed to take neutral and clear and
1)Data Sources" Collecting ": influential data in the process of measuring performance
 Structured interview with expert (IT Department). indicators to solve problems in the data warehouse.
 The data acquisition documents using paper and flash
memories. 3) The Transformation Process
 Note taking. Real data It includes all the data collected from the
 conducted Interviews in person or over the telephone sector used for the measurement process of performance
 Interviews done formally (structured) or semi-structured indicators, it is usually inconsistent, in addition to sometimes
there is a lot of missing and unclear data, so an integration
2) The Data Extraction process is carried out for the data and the missing data is
Through the research paper, the data used in measuring completed and the inconsistent data is excluded, and the
performance indicators were collected within one sectors, and process of changing those data will start to work ERP
the presence of duplicate data will lead to mistakes and low Schema.
performance in the Neural Networks.

IJISRT21OCT210 www.ijisrt.com 297

 Volume 6, Issue 10, October – 2021 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
i. The Missing value: V. TARGET DATA SET
The missing data is deleted or replaced, and in the case
of the replacement process, by setting a numerical rate based The research paper applied 65% of records purposed for
on the existing data, in a mean substitution, the average value coaching the neural network and 35% will maybe be
of a variable is used in place of the missing data value for that implemented as a completely impartial test on the network.
same variable.
Also, that is because of the fact, if the neural network
ii. The Noisy data : has been Implementation for a passing extravagant pair of
The noisy data are data with a large amount of information which were helpful for the training, it will not be
additional meaningless information in it called noise. It also particular it found to "prediction" or to "saving" pattern.
includes any data that a user system cannot understand and
interpret correctly, the error can be in the calculated values, The neural network needs to increase the training
and there are many methods that handle the wrong data such process. Whenever new inputs are introduced into the neural
as clustering, it is able to handle the errors inside the data that network, it trains on it and learns through it. Therefore, it was
resulted from different operations. necessary to perform two operations: the first process, which
is the input process, and the next process is the training
iii. The Inconsistent data: process for the network.
The Inconsistent data is generally exacerbated by data
redundancy. However, it differs from data redundancy in that VI. CONCLUSIONS
it usually points to problems with the content of the database
rather than its design and structure. The aim of the research paper is to present A New
Framework for improving key performance indicators using
There are data inconsistencies in many transactions that Business Intelligence Technics As data mining, which greatly
need to be resolved, There is a lot of existing data that has helps in extracting hidden relationships between data , and
been repaired and modified and prevent the presence of such improving the new key performance indicators that will be
contradictory data within the data warehouse. followed by the institution and try to choose the best indicator
among a group of indicators, and Ensuring successful access
iv. Data Integration: to institutions for their goals (improvement - correct
Data collecting from many sources. Data integration deviations - making decisions) and measuring the success
helps to avoid inconsistencies and improve mining speed and rates of the institution, and Identify the most important
quality. factors affecting decision-making by classifying them by
degree of impact.
This is done by saving all the data in the data warehouse
and using more than one method to access that data in one REFERENCES
place and a several database, Careful integration of data from
multiple resources helps reduce and prevent redundancy and [1]. Abd El-Aziz Ahmed, Hesham A. Hefny, Mohammed
disparity in the data set that they cause. This can help boost Badawy (2018). Exploring and Measuring the Key
accuracy, reliability, and rate from your next mining process. Performance Indicators in Higher Education
Institutions, International Journal of Intelligent
v. Data Selection: Computing & Information Science, Volume18.
Selecting relations of data of given fascination for the [2]. Ahmed Mohamed Abd El-Mongy, Alaa el-Deen
research is the simplest manner to collect very last outcomes Hamouda, Nihal Nounou, Abdel-Moneim A. Wahdan
distinctly applicable to the lookup requirements. (2013), Design of Prediction System for Key
Performance Indicators in Balanced Scorecard.
This paper choose the right variables effect acquisition International Journal of Computer Applications ,
the best KPIs instances as suggestions for improvement, [3]. Alaskar T. and Poulis E. (2015). Business Intelligence
including deferent number of data transactions, Internal Data Capabilities and Implementation Strategies.
Sources and External Data Sources, choosing the factors that International Journal of Global Business 8(1), .
are supported as a result, This selection will improving key [4]. Albert P.C. Chan, Ada P.L. Chan, (2004) "Key
performance indicators. performance indicators for measuring construction
success", Benchmarking: An International Journal, Vol.
vi.Data Transformation: 11.
We find that the data transformed during the [5]. Fink L., Yogev N. and Even A. (2017).Enhancing the
normalization method is very useful for classification Business Value of Business Intelligence: The Role of
algorithms as well as for neural networks. Shared Knowledge and Assimilation. Journal of
Information Systems.
Normalization greatly helps each computed property to [6]. Gonzales, R., Wareham, J., Serida, J. (2015). :
develop the learning process, this method of normalization Measuring the Impact of Data Warehouse and Business
processes all data in appropriate forms to perform data Intelligence on Enterprise Performance in Peru: A
mining. Developing Country. Journal of Global Information
Technology Management. 18 .

IJISRT21OCT210 www.ijisrt.com 298

Volume 6, Issue 10, October – 2021 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
[7]. Gordon S. Linoff, Michael J. Berry(2011): Data Mining [22]. Višnja Istrat, Sanja Stanisavljev, Branko
Techniques: For Marketing, Sales, and Customer Markoski(2015), the role of business intelligence in
Relationship Management. Wiley Computer Publishing; decision process modelin,the European journal of
3rd Ed. applied economics .
[8]. Hanelt, A., B. Hildebrandt and J. Polier (2015). [23]. Vo, Quoc Duy et al (2017). "Next Generation Business
“Uncovering the role of is in business model innovation Intelligence and Analytics." IEEE Communication
– a taxonomy-driven aproach to structure the field.” In: Surveys & Tutorial.
Twenty-Third European Conference on Information [24]. whitepaper(2017),Key Performance Indicators, Six
Systems (ECIS). Münster, Germany. Hartmann. Sigma, and Data Mining, Data Driven Decision Making
[9]. Jesús Peral⁎, Alejandro Maté, Manuel for Financial Institutions, Information Technology
Marco(2017),Application of Data Mining techniques to [25]. Jiawei Han, Micheline Kamber, and Jian Pei: Data
identify relevant Key Performance Indicators. Journal Mining: Concepts and Techniques, 3rd Ed., Morgan
Computers & Industrial Engineering. Kaufmann Publishers, Mar. 2011
[10]. Jiawei Han, Micheline Kamber, and Jian Pei(2011): [26]. Ying Huang, Bing Quan Huang, M. Tahar Kechadi: A
Data Mining: Concepts and Techniques, 3rd Ed., Rule-Based Method for Customer Churn Prediction in
Morgan Kaufmann Publishers. Telecommunication Services. PAKDD 2011: 411-422
[11]. Julia furmankiewicz, Małgorzata furmankiewicz, piotr [27]. Nhien-An Le-Khac, Martin Bue, Michael Whelan, M.
ziuziański(2015),implementation of business Tahar Kechadi: A Clustering-Based Data Reduction for
intelligence performance dashboard for the knowledge Very Large Spatio-Temporal Datasets. ADMA 2010:
management in organization, Computer Publishing. 43-54
[12]. Mari Abe, Jun-Jang Jeng, Yinggang Li (2007), A Tool [28]. S.Xu, M.Zhang, “Data mining - an adaptive neural
Framework for KPI Application Development. IEEE network model for financial analysis” Information
International Conference on e-Business Engineering. Technology and Applications, IEEE, ICITA, pp.336-
[13]. Navin K. Deva, Ravi Shankarb, Rachita Guptac, Jingxin 340, 2005.
Dongd(2017),Multi-criteria evaluation of real-time key [29]. J.Rabunal, J.Dorado, “Artificial neural networks in
performance indicators of supply chain with real-life applications”, pp.297-303, 2006.
consideration of big data architecture, Journal [30]. X.Ni, “Research of Data Mining Based on Neural
Computers & Industrial Engineering. Networks”, World Academy of Science, Engineering
[14]. OlszakC.M.(2016).Toward Better Understanding and and Technology 39, 2008.
Use of Business Intelligence in Organizations. [31]. Y.Singh, A.Singh chauhan, “NEURAL NETWORKS
Information Systems Management . IN DATA MINING”, Journal of Theoretical and
[15]. Popovicˇ A. and Yeoh W. (2016).Approach to Building Applied Information Technology, 2009.
and Implementing Business Intelligence Systems.
Interdisciplinary Journal of Information, Knowledge,
and Management.
[16]. Richards G, Yeoh W, Chong AY-L, et al. (2014) An
empirical study of business intelligence impact on
corporate performance management. PACIS 2014:
Proceedings of the Pacific Asia Conference on
Information Systems 2014. AIS eLiberary.
[17]. Ruhollah Tavallaei, Sajad Shokohyar , Seyedeh Mehrsa
Moosavi , Zahra Sarfi (2015)Assessing the Evaluation
Models of Business Intelligence Maturity and
Presenting an Optimized Model , International Journal
of Management, Accounting and Economics.
[18]. Salmasi, M. K., Talebpour, A., & Homayounvala, E.
(2016). Identification and classification of
organizational level competencies for BI success.
Journal of Intelligence Studies in Business.
[19]. Sangar, A. B., Hesar, Z. E., Asl, M. S. and Tahmores,
K. (2015). Research article proposing IS success models
for measuring business intelligence system (BIS)
success and analytical literature review on BIS
measurement. ANARE Res.
[20]. Stodder, David (2017). Using Design Thinking to
Unleash Creativity in BI and Analytics Development.
Renton, WA: tdwi.
[21]. Stodder, David (2017). Accelerating the Path to Value
with Business Intelligence and Analatics. TDWI.

IJISRT21OCT210 www.ijisrt.com 299