Applying artificial immune system and ant algorithm in air-conditioner market segmentation
Introduction
Recently, data mining has attracted a lot of attention in the information industry and academic research due to the wide availability of large amounts of data and the imminent need for turning such data into useful information and knowledge. The gained information and knowledge can be used for applications in business management, production and operation management, forecasting, marketing analysis and science exploration. In the domain of data mining and knowledge discovery, the common issues include association rule mining, classification, prediction and clustering analysis. With the trend of data enormously increasing, how to deal with these huge data for quickly data mining and knowledge discovery becomes one of the most important issues. Data clustering groups the collected data with similar characteristics into the same cluster and analyzes the relationship among these objects or data. Data clustering is not only a single data mining method but also a pre-process of data mining, knowledge discovery and data collection.
Ant algorithm is a heuristic method proposed to solve optimization problems. In a recent research of clustering problem, Ant system-based clustering algorithm, ASCA (Kuo, Wang, Hu, & Chou, 2005), applied the ant algorithm to merge data into the same cluster if they are within the distance of some specified standard deviation from the centroid. However, the objects near the edge of each cluster may result in dissimilarity with higher total within cluster variance (TWCV) than the objects gathered around the centroid. In this situation, it is more rational to group these data into another independent cluster which reduces the TWCV of both clusters. In this paper, we proposed the immunity-based ant clustering algorithm, IACA, which integrates the ant algorithm and the artificial immune system to resolve such situation and search for optimization of clustering problems.
Section snippets
Clustering analysis
Clustering is the process of grouping a set of abstract or physical objects into classes composed of similar objects. It divides data into several groups. The purpose of clustering analysis is to find the difference among each groups and the similarity in the same group. A group is also called a cluster. The cluster is also a collection of data objects similar to one another in the same cluster and dissimilar to others in other clusters.
Unlike classification, clustering does not know the
Definitions
The following definitions are adopted from Kuo et al. (2005). for the further development of our proposed clustering algorithms:
- •
Let E = {O1, O2, … On} be the data set with n objects, where O indicates the objects (or data, items) collected from the database. And each object has k attributes (dimensions), where k > 0. The data format is shown as in Table 3.1.
- •
n: the number of objects.
- •
nc: the number of clusters.
- •
T: the set composed of the used objects.
- •
Tk: the set T named k.
- •
D(i,j): the Euclidean distance
Case study
In order to further examine the proposed algorithm, a real-world case data is applied. In this case, we investigate questionnaires of the customer satisfaction and perception to a large 3C appliance chain store. The proposed clustering method is implemented to group customers into different segments for further customer relation management.
Conclusion
In this paper, we proposed the immunity-based ant clustering algorithm, IACA. It consists of two types of search algorithms, the ant algorithms and the artificial immune system. Ant algorithm is utilized to generate the initial solution. Then, the artificial immune system is adopted to search for optimization of clustering problems. In order to strengthen the demonstration of the proposed algorithm, our proposed algorithm is applied to real-world data for air-conditioner market segmentation.
References (26)
- et al.
Ant algorithms and stigmergy
Future Generation Computer Systems
(2000) - et al.
Ant colonies for the traveling salesman problem
BioSystems
(1997) - et al.
Integration of self-organizing feature maps neural network and genetic K-Means algorithm for market segmentation
Expert Systems with Applications
(2006) - et al.
Integration of self-organizing feature map and K-Means algorithm for market segmentation
International Journal of Computers and Operations Research
(2002) - et al.
Application of Ant K-Means on clustering analysis
Computers and Mathematics with Applications
(2005) - et al.
An immunity-based ant colony optimization algorithm for solving weapon-target assignment problem
Applied Soft Computing Journal
(2002) - et al.
A tabu-search-based heuristic for clustering
Pattern Recognition
(2000) - et al.
A genetic approach to the automatic clustering problem
Pattern Recognition
(2001) - Berkhin, P. (2002). Survey of Clustering Data Mining Techniques, Accrue Software, Inc....
- et al.
Distributed optimization by ant colonies
Immunity-based systems: A survey
Proceeding of the IEEE Transactions on Systems, Man and Cybernetics
Cluster merging and splitting in hierarchical clustering algorithms
IEEE International Conference
The ant colony optimization meta-heuristic
Cited by (19)
A new oversampling method and improved radial basis function classifier for customer consumption behavior prediction
2022, Expert Systems with ApplicationsCitation Excerpt :IA is usually used to find a global optimal solution that has a good balance between population diversity and convergence, and restrains the immature convergence (Duma & Twala, 2019). An antigen is defined as the objective function of the actual problem, and an antibody can be regarded as a candidate solution to the actual problem (Chiu et al., 2009; Samigulina & Samigulina, 2019). Due to the diversities of antibodies, IA can prevent the RBFNN model from falling into a local optimum in determining the RBFNN centers (Khan et al., 2020).
Evolutionary computing applied to customer relationship management: A survey
2016, Engineering Applications of Artificial IntelligenceCitation Excerpt :Ant algorithm is utilized to generate good solutions for clustering while AIS is used for the optimization of clusters. Chiu et al. (2009) generated segments using the immunity-based ant clustering algorithm which is an integration of an ant algorithm and artificial immune system. Data from a large 3C appliance chain store was used.
A novel fuzzy hybrid quantum artificial immune clustering algorithm based on cloud model
2014, Engineering Applications of Artificial IntelligenceCitation Excerpt :Evolutionary algorithms may resolve some of the issues associated with non-Gaussian distribution; the solutions provided by evolutionary algorithms may get trapped in local optimum. A hybrid approach using evolutionary algorithms was proposed to improve the accuracy of linear discriminate analysis (Chiu et al., 2009). Recently, the human immune systems had aroused researcher׳s interest due to it useful mechanisms which can be used and exploited for information processing in a complex cognition system.
Optimum design of hybrid renewable energy systems: Overview of different approaches
2012, Renewable and Sustainable Energy ReviewsCitation Excerpt :Best paths are next to the food since many more ants are next to good and rich places and all together they contribute to increase the pheromone level of those paths leading to the food. The best path (it can also be called as “shortest path”) towards the food in the proposed application is the highest possible value of performance index in an optimization procedure [207]. A block diagram of the ACO approach that can be helpful for upcoming researchers focused on hybrid system sizing is shown in Fig. 5 [200].
A novel hybrid immune algorithm and its convergence based on the steepest descent algorithm
2011, Applied Mathematics and ComputationCitation Excerpt :Zhang et al. [20] provided a dynamic threshold AIA to improve the convergence speed and to prevent premature convergence. Chiu et al. [21] proposed the immunity-based ant clustering algorithm. Tsai et al. [22] developed a modified Taguchi immune algorithm based on both the features of an artificial immune system and the systematic reasoning ability of the Taguchi method.
Solving a capacitated fixed-charge transportation problem by artificial immune and genetic algorithms with a Prüfer number representation
2011, Expert Systems with ApplicationsCitation Excerpt :To be protective, the immune system must learn to distinguish between its own cells and malefic external invaders. This process is called self/non-self discrimination (Chui, Kuo, & Lin, 2009). Vertebrates possess an adaptive immune system that can learn to recognize, eliminate and remember specific new antigens.