A Study and Survey of Artificial Immune Systems | Open Access Journals

ISSN ONLINE(2320-9801) PRINT (2320-9798)

A Study and Survey of Artificial Immune Systems

Dr.Kathir. Viswalingam1 and Ayyappan2
  1. Dean (R&D), Bharath University, Chennai, India
  2. Assistant Professor, Department of Information Technology, Bharath University, Chennai, India
Related article at Pubmed, Scholar Google

Visit for more related articles at International Journal of Innovative Research in Computer and Communication Engineering


Artificial immune systems (AIS) square measure a category of computationally intelligent systems that take into account several properties of natural immune system .Several AIS square measure wide used in completely different application areas such as classification, clustering, web mining, virus detection, learning, image process, AI management, bio-informatics and anomaly detection. Among this classification and bunch square measure wide used areas. Most of the bogus system employed in the classification and bunch space build use some key options of AIS like feature extraction, recognition and learning. This paper provides a good survey regarding artificial immune systems that square measure employed in the classification and bunch space sand conjointly build use of the options like feature choice, pattern recognition and machine learning.


Artificial immune System; Classification; Clustering; Feature Selection; First Staten Machine Learning.


The artificial immune systems (AIS) square measure used patterns and build call supported the class of the for modelling the natural immune system [5] that comes underneath the category of computationally intelligent system. Properties of AIS embrace lustiness, ability, diversity, turnover of elements and autonomy. There square measure many application square measure as are there for AIS [7] [5] which may be summarized as follows, classification and bunch, anomaly detection, laptop security, learning, image process ,robotics ,virus detection and net mining. Classification/clustering square measure most wide used application space of AIS. Classification and bunch needs feature choice, recognition and learning, that square measure the key options of artificial system. bunch is a manner to mechanically derive broad patterns or structures from a corpus generally referred to as Associate in Nursing unsupervised learning approach. Classification is the step of mapping object to notable categories whose facet effects is bunch, it need derivation of categories or coaching a classifier . Classification and bunch build use of feature choice, recognition and learning. Feature choice is to search out the minimally sized feature set that’s necessary and enough to the target idea. For feature choice we have a tendency to build positive that classification accuracy doesn't considerably decrease [10].clustering uses in the main three models like filter, wrapper and hybrid model for feature choice .the goal of pattern recognition (PR) is that the classification of knowledge, object or patterns into classes. For this purpose it examine however machines will extract data, learn to acknowledge pattern [1].different classification technique and cluster analysis square measure used for recognition. Machine learning target finding patterns from quality of information returning from a supply of interest. Feature based mostly} and similarity based approaches square measure 2 main machine learning techniques. Machine learning employed in each classification and bunch application in numerous ways in which .some application provide the definition of machine learning because the ability of a machine to enhance it document classification performance [6]. The paper is organized as follows: section a pair of presents regarding the artificial immune organisation. Section three discusses the artificial immune bunch systems. The conclusion of this paper is given in section four.



Classification is the most wide used application space of AIS. Artificial immune organization gift some artificial immune system that comes underneath this space and conjointly build use of the options like feature choice , pattern recognition and machine learning of AIS,


There square measure many paper bestowed for AIS, that support the classification with the facilitate of feature choice. Some examples square measure given as follows.


By the behavior of immune system a new classifier is contemplated. Native feature choice [7] is achieved and known as exclusive specialty of this classifier. The combining of protein evolves Associate in nursing epitope and a paratope with substance determined in amino acid residues. Primarily based on this behavior feature choice is originated. The energetic residues participate in the interaction. All the organism choice section protein receptors square measure obtained. The memory set is made once the protein binding with substance. The memory set is made once the protein [identifying] and substance square measure associated along. The memory set will be reduced once sure instances square measure omitted. to cut back the information native feature choice and omission of instances square measure undertaken. Process time reduces once there square measure fewer options to classify. solely a 100 percent of information was necessary to classify, charity ninetieth of information was reduced that is necessary for classification The classifier is made with 2 commonplace user outlined parameters which is able to limit the world and native characteristics of feature house looking out. Crosschecking was created on the classifier to acquire correct results on the existing problems. The K-Nearest Neighboring algorithmic rule, support vector machines and Random forest classifier square measure used for testing. Analyzing the results its discovered that the accomplishment of the projected classifier is manufacturing correct results once compared to different immune impressed classifier and different general classifiers.


Tournament looking out technique [6] is introduced for feature choice downside. Simulated tempering or generic algorithmic rule square measure already existing ways for feature choice, such ways square measure terribly sophisticated. The benefits of projected ways square measure it's easier to use and less sophisticated. TSM for FS is a straightforward random looking out technique. This technique uses one parameter for dominant either native or world looking out properties of the algorithmic rule.


The main objective of this paper is to style a pattern recognition engine and its supported ideas derived from class immune systems. The planning of immune-81 [8] is predicated on code abstractions of T cells, antibodies, and their interactions. Creation of B cells population is controlled by the T cells. The b cells contend for recognition of ‘unknowns’. The measurements for Immunos-81 square measure, 2 commonplace machine learning information sets, were used to check the popularity capabilities of Immunos-81.The initial set(Cleveland), consisting of various cases of patients. It is used to perform a ten-way cross-validation. Once finishing the validation runs, the Cleveland dataset was used as a coaching set. This coaching set is used for presentation of the second information set, consisting of multiple unknown cases. System offers a viable paradigm for the planning of pattern recognition systems.immune- 81 could be a variety of supervised learning system. Compared to different machine learning algorithms Immune-81 performed well.


There square measure variety of harm detection and localization ways are projected. A number of them explore the structure harm classification downside. This paper presents a synthetic Immune Pattern Recognition (AIPR) [4] model for the harm classification in structures, which includes many characteristics of the natural immune system. Mimicking immune recognition is employed to attain structure harm pattern recognition. The main options of mimicking immune recognition mechanism square measure adaptation, evolution, and immune learning. Set of detector nodes square measure wont to collect time series information and structure harm classification is predicated on this series of knowledge. The AIPR structure harm classifier (AIPR-SDC) consists of 2 major stages. Initial stage is that the information pre-processing and have extraction. During this stage all detector information is standardized and feature vectors square measure generated, memory cell set and protein set for all the categories also are initialized within the initial stage. within the second stage, the coaching substance stimulates the protein set and therefore causes some of antibodies to provide clones.


Machine learning is additionally employed in classification so as to develop AIS (Artificial Immune System).


The objective of AIS for Intrusion Detection [12] is to cut back the occurrences of intrusion detection in laptop networks. The work achieves this through group action artificial algorithmic rule with pattern recognition and machine learning. Such a system is developed supported the danger theory model of human system (HIS). The danger model and its application square measure helpful for the activation of malicious behavior defense that successively to make a totally decentralised model. The nerve fibre cell behavior and T-cell mechanisms square measure the mechanisms used in this paper. Such systems square measure wide used in autonomous arms.


In this paper a classifier supported the machine learning technique is employed to strain spam emails mechanically. This technique introduce a reliable antispam filters. Completely different machine learning techniques square measure introduced for email classification. Naive Bayes classifier technique, K-nearest neighbor classifierMethod, Artificial Neural Networksclassifier technique, Support Vector Machines classifier technique, Artificial Immune Systemclassifier technique and Rough sets classifier technique square measure most typically used ways. Naive Bayes classifier technique is most economical compared to different higher than mentioned classifier ways.


Artificial immune bunch system tell regarding however the key options of AIS like feature choice, pattern recognition and machine learning facilitate in many AIS, that square measure comes underneath the bunch application space,


There square measure many paper bestowed for AIS, that support the bunch with the facilitate of feature choice. Some examples square measure given as follows.


This paper [13] aim to deal with the impact of feature reduction in anomaly intrusion detection system. It conjointly discover bunch attack with the use of bio- impressed artificial immune network algorithmic rule, this algorithmic rule is used on the reduced dataset. One main advantage of AIS for AID is lustiness since it selects solely the foremost vital options rather than choosing all the options. Different blessings square measure detection accuracy and reduction of false alarms. This paper is viable when put next with k-means bunch technique.


It introduces a fuzzy linking approach for shouting image information set. The retrieval of pictures from shouting information set achieved through feature extraction and similarity criterion. CBIR system permits retrieval of pictures from an info with improved noise tolerance compared to different systems. The effectiveness of mage retrieval will be increased by applying artificial immune bunch algorithmic rule. Lustiness is one in all the most blessings of CBIR systems. Classic bar graph is one in all the CBIR systems that have the capability of resolution complicated issues. Content primarily based image retrieval performs higher than histogram-based technique. Several papers were introduced in AIS that support with the facilitate of pattern recognition.


The objective of INEPR is presents a process model for the nascent pattern recognition, supported the immune network theory and hierarchical bunch algorithms. Dynamic production of internal image for the input file patterns facilitate for the nascent pattern recognition. hierarchical bunch algorithms accountable for creation of Associate in Nursing protein memory cell clustering; such protein memory cells square measure fashioned from the members of internal image. Determination of variety of clusters employed in the memory cell bunch could be a difficult task. Previous papers have to be compelled to calculate it ahead however INEPR will mechanically calculate the simplest variety of clusters. Analysis of graph and L technique square measure introduced in INEPR to estimate the quantity of clusters.


The objective of Hybrid Artificial Immune Pattern Recognition (HAIPR) is introduced Associate in Nursing unsupervised structural harm pattern recognition supported the fuzzy bunch and therefore the artificial immune pattern recognition (AIPR). The inaccessibility of pattern data of the coaching information is one in all the constraints of unsupervised structural harm pattern recognition. The implementation of fuzzy bunch facilitate to beat such inaccessibility of coaching information. Quality of memory cell will be improved by artificial immune pattern recognition technique primarily based on Associate in Nursing immune learning mechanism. Artificial immune pattern recognition technique is utilized to enhance the standard of memory cells for every harm pattern. The HAIPR for the unsupervised harm pattern recognition victimization the IASC—ASCE benchmark structure is higher than the present algorithms such as FC-SVM and FC-Naïve Bayes


This section describe regarding however machine learning feature of AIS is used in artificial immune systems, that comes underneath the application space bunch. Some examples square measure given as follows. The aiNet is one amongst the AIS algorithmic rule that will be with success used in several machine learning tasks. The aiNet exploiting the biologically- impressed options of the immune system and it conjointly perform well on elementary bunch tasks. This paper proposes the utilization of the aiNet to a lot of complicated task of document bunch. The aiNet perform Associate in Nursing method|biological process} process on the data supported immune network and affinity maturation principles. It can end in reduced information redundancy and retrieve smart bunch results. So as to cut back the time quality principal part analysis is integrated into this technique. Finally the results square measure compared with hierarchical clustered bunch and k-means, that square measure some classical document bunch ways.


In this model a kernel primarily based approach [9] was introduced. In this approach represent antibodies as linear combination of antigens. This enables the definition of a mutation mechanism while not presumptuous a vector illustration of antibodies. The model assumes Associate in nursing affinity live between antigens. {This idea|this idea|this idea} is supported by form house concept and gift in current models. Some tasks for future work square measure i) to check the utilization of the kernel trick with completely different varieties of cluster structures ii) introduce a plan referred to as protein saturation.


In this paper I created a study regarding artificial system (AIS) and that i found completely different application areas of AIS. A number of them embrace classification and bunch, optimisation, learning, image process, AI etc. Among these classifications and bunch is most generally used, therefore I created a study regarding completely different papers, that square measure used in this space. In these papers classification and bunch build use of necessary options of AIS such as feature choice, pattern recognition and machine learning.


  1. Bo bird genus ,ChuanzhiZang , (2011), “A hybrid immune model for unsupervised structural harm pattern recognition” , professionalSystems with Applications thirty eight (2011) 1650–1658.

  2. Bo bird genus and ChuanzhiZang (2011), “Emergent harm pattern recognition victimization immune network theory” good Structures andSystems, Vol. 8, No. one (2011) 69-92.

  3. Bo bird genus, ChuanzhiZang (2009), “Artificial immune pattern recognition for structure harm classification”, Computers and Structureseighty seven (2009) 1394–1407.

  4. Emma Hart and eating apple Timmis, user interface Wei-Hua (2005), “Application Areas of AIS: The Past, this and therefore the Future”,ICARIS LNCS 3627, pp. 483–497, 2005.

  5. GrzegorzDudek (2010), “Tournament looking out technique to Feature choice Problem”, ICAISC half ii, Pg: 437-444, Springer-verlag berlinHeidelberg 2010.

  6. GrzegorzDudek (2012), “An Artificial system for Classification with native Feature Selection”, IEEE Transactions On biological processComputation, Vol. 16, No. 6, Gregorian calendar month 2012.

  7. Theologist H. CARTER, MD(2000), “The Immune System as a Model for Pattern Recognition and Classification”, Journal of the yankMedical information science Association Volume seven #1 January / Feb 2000.

  8. Juan C. Galeano, Ing., Fabio A. Gonzále (2008),”Towards a KernelBased Model for Artificial Immune Networks”, RevistaAvances nutSistemas e Informática, Vol.5 No. 1, Edición exceptional, Medellín, dressing First State 2008, ISSN 16577663.

  9. M. Dash, H. Liu (1997), “Feature choice for Classification”, Intelligent information Analysis one (1997) 131–156.

  10. MortezaAnaloui, and MaedehBeheshti (2011), “Content-based Image Retrieval victimization Artificial Immune System (AIS)Clustering”, IMECS,Vol. I, March 16-18, 2011.

  11. MunaElsadig Mohamed, BrahimBelhaouari Samir, Azween Abdullah (2010), “Immune Multi agent System for Network IntrusionDetection victimization Non-linear Classification Algorithm”, International Journal of laptop Applications (0975 – 8887) Volume 12– No.7,Gregorian calendar month 2010.

  12. MuradAbdoRassam ,Mohd. AizainiMaar of (2012), “Artificial Immune Network bunch approach for Anomaly Intrusion Detection”,Journal Of Advances In data Technology, Vol. 3, No. 3, August 2012.

  13. Na Tang , V. RaoVemuri (2005), “An Artificial Immune System Approach to Document Clustering”, SAC’05, March 13–17, 2005, SantaFe, New Mexico, USA.