SCHOOLOF MANAGEMENT2017-2019BATCH ASSIGNMENT LITERATURE REVIEWSubmitted by: Namitha sudhakaranRoll: 178920Subject: business research methodologySubmitted to: Dr.
Ritanjali MajhiDate of submission: 28th of january, 2018Running Head: artificial neural networks in business Literaturereview Namithasudhakaran National Institute of Technology, Warangal ABSTRACTArtificial neural networks are commonly used inbusiness but the studies and finding regarding that is very few in number. Theyare connected nodes or units and each unit passes a signal through it. Here i am trying to gothrough the main functional areas and uses of artificial neural networks byreviewing the literature papers on the topic. In the past decade it grow up andperforming many activities in wide variety of areas. reviewed more than 6papers to find out different functional areas and usages of artificial neuralnetwork. INTRODUCTIONUtilization of neural networks are largely increasedin the two decades. Artificial neural networks are computational structuresthat are used to emulate the knowledge in the central nervous system.
Here I amtrying to analyze the recent papers in artificial neural network to prepare aliterature review on the papers that are already presented. i had gone throughsome papers and trying to convey the main ideas of the particular one. Theyhave high efficiency and easily adaptable to use in different kind of analysis.Most applications of this can be published in bankruptcy prediction and stock forecasting. Most common researcharea of artificial neural network should be come under finance in future. All the studies revealing the importance of this artificial intelligencemethod and illustrate about recentresearch for both academics and practitioners.
This review paper not onlyemphasizes the historical progressions in the field of neural networks, itdiscusses the prospective development in the neural network research areas. Paper1: ArtificialNeural Networks: State of the Art in Business IntelligenceAuthor :Sunil Sapra.Departmentof Economics and Statistics, California State University, Los Angeles, CA90032, USAThe paper demonstrating about how ANN is used forbusiness and the important of ANN in business forecasting.ANNs are an excellenttool for forecasting, but their results are difficult to interpret since ANNsintroduce complex interactions. In the absence of appropriate controls, theANNs can over-fit the data producing overoptimistic predictions. ANNs work verywell for large complex data sets in comparison with statistical methods. A keyweakness of the ANNs is that they do notpossess sound statistical theory for inference, diagnostics, and modelselection. ANNs used carefully, can outperform statistical methods for certainproblems.
in some areas ANN failed . things have changed over the past fewyears due to the feasibility of deep networks made by new training techniques, availability ofbillions of documents, images and videos available for training purposes withthe rise of internet, and the realization that graphical processing units(GPUs), the specialized chips used in PCs and videogame help to generategraphics, that are suited to modeling neural networks. Along With deepernetworks, more training data and powerful and new hardware, also deep neuralnetworks that have made rapid progress in the areas of speech recognition andimage classification and language translation. now it cover all areas and performing up to the expectationcompared to previous.PAPER 2: Deeplearning in neural networks: An overviewAUTHOR: jurgen schmidhuber This paper dealing with historical survey of usageand popularity of artificial neural network In recent years, deep artificialneural networks have won numerous contests in pattern recognition and machinelearning.
This historical survey summarizesrelevant work, this is also comment on Deep Learners are distinguished by thedepth of credit assignment paths which are chains of possibly learnable, normallinks between actions and effects. deep supervised learning , unsupervised learning,reinforcement learning,evolutionary computation, and indirect search for shortprograms and also encoding deep and large networks. and their usage ondifferent historical survey areas.Paper 2: Improved system identification usingartificial neural networks and analysis of individual differences in responsesof an identified neuronAuthor: AliciaCo stalago Meruelo David M.Simpson Sandor M.Veres Philip L NewlanThis paper deals with the modeling and process ofartificial neural networks mathematical modeling is used to understand thecoding properties and dynamics of responses of neurons and neuralnetworks.
analyze and evaluate theeffectiveness of Artificial Neural Networks (ANNs) as a modeling tool for motorneuron responses. ANNs used to model the synaptic responses of anidentified motor neuron and also, the fast extensor motor neuron, one of thedesert locust in response to displacement of a sensory organ, the femoral chordtonal organ, this monitors movements ofthe tibia relative to the femur of the leg which is under consideration. Theaim of the study are, to determine the potential value of ANNs as tools tomodel and investigate neural networks, and second to understand the main general propertiesof ANNs across individuals and to different input signals and last, tounderstand individual differences in responses of the identified neuron. Theperformance of the models are generated by the ANNs was compared with thosegenerated through the previous mathematical models of the same neuron. then Theresults suggest that ANNs are significantly good compared to LNL and Wienermodels in predicting specific neural responses to Gaussian White Noise, but notsignificantly different when tested with sinusoidal inputs. They are also ableto predict responses of the same neuron in different individuals irrespectiveof which animal was used to develop the model, although notable differencesbetween some individuals were evident. this paper is all about the applicationof ANN in the medical industry. Paper3: Failure load prediction of single lapadhesive joints using artificial neural networksThe objective of this paper was to predict thefailure load in single lap adhesive joints subjected to tensile loading byusing artificial neural networks.
Experimental data obtained single lapadhesive joints with various geometric models under the tensile loading. Thedata are arranged in a format such that 2 input parameters cover the length andwidth of bond area in single lap adhesive joints and also the correspondingoutput is the ultimate failure load. An artificial neural network model wasdeveloped to estimate relationship between failure loads by using geometricdimensions of bond area which as input data. A 3-layer feed forward artificialneural network that utilized a particular kind of algorithm model its used inorder to train the network. It is observed that artificial neural network modelwhich can also estimate failure load of single lap adhesive joints withacceptable error. and the results showed that the artificial neural network isan efficient alternative method to predict the failure load of single lapadhesive joints. Papre4: Artificial neural networks inbusiness: Two decades of researchAuthor : MichalTká? Robert vernorThis paper dealings with the research studies andprogress that had already happened in the area of artificial neural network andwhat are the areas and upcoming trends that the artificial intelligence arecovered when the business trends are developed more and more, this paperinclude the literature review of the papers which are already done in the field of ANN artificial neuralnetworks have been extensively used in many business applications. Despite thegrowing number of research papers, only few studies have been presentedfocusing on the overview of published findings in this important and populararea.
Moreover, the majority of these reviews were introduced more than 15years ago. The aim of this work is to expand the range of earlier surveys andprovide a systematic overview of neural network applications. The authorcovered a total of 412 articles and classified them according to the year ofpublication, application area, type of neural network, learning algorithm,benchmark method, citations and journal. Our investigation revealed that mostof the research has aimed at financial distress and bankruptcy problems, stockprice forecasting & decision support, special attention to classificationtasks. Besides conventional multilayer feed forward network with gradientdescent back propagation, also various hybrid networks has been developed inorder to improve the performance of standard models. Even if neural networkshas been established as well-known method in business, there is enormous spacefor additional research in order to improve their functioning and increase ourunderstanding of this influential area of the subject. Paper5: Studying the Effect of ActivationFunction on Classification Accuracy Using Deep Artificial Neural NetworksAuthor: Serwa A ,Faculty ofEngineering in El- MatariaArtificial Neural Networks (ANN) is widely usedin remote sensing applications. Optimizing ANN still an enigmatic field ofresearch especially in remote sensing.
This reaserch work is a trial todiscover the ANN activation function to be used perfectly in classification thefirst step is preparing the reference map then assume a selected activationfunction and receive the ANN output. Thelast step is comparing the output with the reference to reach the accuracyassessment. The research result is fixing the activation function that isperfect to be used in remote sensing classification. A real multi-spectralLandsat 7 satellite images were used and was classified and the accuracy of theclassification was assessed with different activation functions. The sigmoidfunction was found to be the best activation function. and the entire paper isdealing with the different kind and area of application of ANN in remotesensing. Paper6 : ArtificialNeural Networks Controller for Crude Oil Distillation Column of Baiji RefineryAuthors :Duraid FadhilAhmed and Ali Hussein KhalafThis paper is dealing with a specific application ofANN in the area of crude oil distillation column of a particular refinery andhow the process are going on there with the help of ANN.A neural networkscontroller is developed and used to regulate the temperatures in a crude oildistillation unit.
Two types of neural networks are used; neural networkspredictive and nonlinear autoregressive moving average (NARMA-L2) controllers. Theneural networks controller that is implemented in the neural network toolboxsoftware uses a neural network model of a nonlinear plant to predict futureplant performance. Artificial neural network in MATLAB simulator is used tomodel Baiji crude oil distillation unit based on data generated fromaspen-HYSYS simulator.
A comparison has been made between two methods to testthe effectiveness and performance of the responses. The results show that agood improvement is achieved when the NARMA-L2 controller is used. Also shownpriority of neural networks NARMA-L2 controller which gives less offset valueand the temperature response reach the steady state value in less time withlower over-shoot compared with neural networks predictive controller. Paper 7: applications of artificial neuralnetworks for medical diagnostics and prognosticsJaouher Ben Ali University of Sousse, TunisiaApplication of ANN in the medical field is alreadydiscussed previously since it is one or other way related with business aspectsagain we have to go to in detail on the another application In the medicalfield, diagnostic and prognostic remain the most important step to identifydisease type and thereby define the adequate treatment before reachingcatastrophic and fatal states.
However, clinical symptoms and syndromes are notsufficient to detect some diseases. Consequently, the definition of newadvanced techniques for medical diagnostics and prognostics are becoming ofgreat interest to assist specialists in clinical researches and hence to ensuresafety for millions of people. Artificial neural networks (ANNs) are inspiredby the way that the brain performs computations: they are classified as one ofthe best and most used soft computing techniques.
In this context, two innovativemethods for early-stage Alzheimer’s disease diagnosis and blood glucose levelprediction of Type 1 diabetes prediction and other cancer image analysis willbe presented, as well as the result interpretation and some case studies. Theaim of this work is to show the great assistance provided by these advancedtechniques to the medical staff where the big data are processed through atrained ANNs leading accurate statistics leading suitable diagnostic decisionmaking CONCLUSIONArtificial neural networks have been taken anenormous attention in last two decades. Much of the research has focused onvarious business disciplines, however, only a small number of surveys have beenpublished in this area. Presented paper has examined 412 neural networkapplications in different areas of business published between 1994 and 2015 inwell-known influential journals.Proper integration of met heuristic methods into theneural network methodology might be a key for achieving the optimalperformance. In general, neural networks have been successfully applied in widerange of business tasks and were able to detect complex and nonlinearrelationships without requiring any specific assumptions about the distributionor characteristics of the data.
REFERANCE· ArtificialNeural Networks: State of the Art in Business Intelligence Sunil Sapra.Department of Economicsand Statistics, California State University, Los Angeles, CA 90032, USA· ofartificial neural networks for medical diagnostics and prognostics· Artificialneural networks in business: Two decades of research Author : MichalTká? Robert vernor· Deeplearning in neural networks: An overview AUTHOR: jurgen schmidhuber · Improvedsystem identification using artificial neural networks and analysis ofindividual differences in responses of an identified neuron Author: Alicia Co stalago Meruelo David M.Simpson Sandor M.Veres Philip L Newlan · Artificialneural networks in business: Two decades of research Michal Tká?c1, RobertVerner? University of Economics in Bratislava, Department of QuantitativeMethods, Tajovského 13, 04013 Ko?sice, Slovakia