Data Mining Process and Algorithms - Essay Example

? Data Mining School a). Using predictive analytics to comprehend behaviors: To remain competitive within the market, sellers need to comprehend present consumer behavior as well as predict those of the future. The accurate prediction of these behaviors and the full understanding of customer behaviors can help them in improving sales, retaining customers, and extending the relationships sustained with customers. This is realized through predictive analysis data mining, which offers the users, impactful insights throughout the organization (Greene, 2012). Predictive analytics is where statistics and mathematics integrate to business and marketing to establish patterns in data and extrapolating the patterns to future business cases and issues, so as to reduce costs, improve response rates, increase the efficiency of processes and consequently boost revenue levels. Data mining has different components, but the most significant is defining the problem, evaluating the available data and developing predictive models. (b). Associations discovery for the commodities sold to consumers helps the retailer or other business to capture the unique identifier of a given product. Through capturing this information, the seller is able to analyze the data, so that they can learn the purchasing behavior of their customers. The information derived is used to support business-related strategies and applications like inventory management, marketing promotions and customer relations management. (c). Mining information on web usage is very important to the effective management of websites, planning the development of adaptive websites, administering business and support services, increasing personalization as well as analyzing the flow of network traffic. Further, fast business growth of businesses forces businesses and customers to face a different situation, where competition plays a major role in determining the strategies adopted by businesses (Greene, 2012). On the other hand, the customer is exposed to more options to choose from, therefore, will need to follow the businesses that depict more value. For example, through discovering that many customers of a given business come from teen customers, may help the company to adjust their targeting outlook, to ensure that it targets the focus group better. (d). Clustering analysis traces groups of data entities or objects that are similar in certain aspects. The members of the different groups are supposed to be more similar to other members, and different from the members of other clusters. The target of clustering is the discovery of high-quality groups, where inter-cluster similarity is lowest but intra-cluster similarity is highest. Through establishing the highest inter-cluster similarity, the characteristics of the members are used or viewed as the customer information that can be tracked or targeted to increase the impact of the business, among the given high-quality cluster. 2. The reliability of data mining algorithms can be done through the validation of data mining modes. The process involves the assessment of the performance of the mining models against real data. This is done through understanding the characteristics and the quality of the algorithms before deploying them into the production environment (Chung, & Gray, 1999). To determine the reliability of data mining algorithms, the deployment of different statistical validity measures is checked, towards determining whether there are issues in the model or the data. The reliability of data mining algorithms is determined through the scalability of the clustering techniques. This is particularly true, in the case of large data sets, where space and speed are high. For example – in the case that the algorithm –in the case of a database that contains millions of records, shows linear or close to linear time complexity, which demonstrates that the reliability of the algorithm is high. The reliability of the algorithm can be determined through the application of tests to determine whether a given point belongs to the cluster in question. In such a case, a reliable algorithm will detect outliers and noise, and eliminate their negative effects or delete them entirely. Through this process, the testing of the validity of the algorithm can be done while the process of clustering is ongoing, or it can be done at the post-processing phase. Following the ability of algorithms to sort the mass digital storage collected using computer systems, since their advent, provided that the reliability of these models is high, then their trustworthiness is not questionable. For example, to overcome the problem of unpredictability in handling data, the development of structured database management mechanisms and databases has increased the ability of algorithms to offer factual information (Han & Kamber, 2000). Therefore, data mining algorithms, considering the established nature of database management systems, offers an effective and trustworthy way of providing assorted information and data. The errors that are likely to arise from the usage of data mining algorithms include querying the wrong question, the failure to test the reasonable nature of the outcomes, ignoring differences in the data, building overly complex models, over generalizing the results and the employment of a single data analysis (Zhou, Jiang & Chen, 2003). 3. (a). The first privacy concern is whether the information collected for the purpose in question will be used for secondary or additional purposes in the future and whether the protection of privacy is inbuilt into the systems at the stage of development and whether the information will be further mined after the collection. The other concern is whether there is a possibility that it will be combined with other information from private or government agencies. (b). The first privacy concern is valid, because the customer must express their concerns that the information collected may be sold or communicated to other agencies, for instance marketers, who will pursue business from them (GAO, 2004). The second privacy issue is valid because the information of the customer may be diverted to other agents that may use it, prior to the checking of privacy. For example, a customer’s private information may be collected by government agencies investigating transactions that appear suspicious. The third privacy concern is valid because combining the information collected with that available at public or private enterprises can lead to harassment or the undue investigation of customers due to possible prototyping of the private information. For example, the collection of an Arab’s names and other information may be subjected to investigation, during the course of investigating terrorist activities and dealings (GAO, 2004). (c). The first privacy concern is addressed through the collection of only vital information, so that the customer can be guaranteed that the information will not be used for other purposes in the future. The second privacy concern is addressed through integrating the data privacy system into the data collection process, right from the development stage. This ensures that the system does not leak any information. The third concern is addressed through ensuring that customers’ information is stored safely and in a way that ensures that it does not infiltrate to others who may need it. 4. In the case of hospitals, predictive analysis has been used to point out the side effects of medication, towards improving the effects of medical practice. The effectiveness of this business strategy is that the hospital acquires more refined information on the treatment types that are most effective for certain diseases. Street clothing lines at Paris situate cameras at the streets, from where they can observe the fashion trends of the time (Siegel, 2013). Following the strategy, the designers recognize the color, fashion trends and the designs that are coming in and those that are losing the market, from the historical data. The patterns drawn are used to predict the next breakthrough in fashion designing. Child welfare organizations are using predictive analysis to minimize the dropout of children from foster parent’s care, because it increases their risk of falling out of the program as a whole. They do this through using a predictive model to match the parents, depending on age, income, ethnicity and education among other variables, and the children that they can match better. The effectiveness of the strategy is that it has improved the success of foster parenthood, by increasing the duration of stay and the overall outcomes of the service for society, the government and the children (Siegel, 2013). References Chung, M., & Gray, P. (1999). Special Section: Data Mining. Journal of Management Information Systems, 16(1), 11-16. Greene, W. (2012). Econometric Analysis, 7th Ed. London: Prentice Hall. Han, J., & Kamber, M. (2000). Data Mining: Concepts and Techniques. Massachusetts: Morgan Kaufmann. Siegel, E. (2013). Predictive Analytics: The Power to Predict Who Will Click, Buy, Lie, or Die. New York: John Wiley. U.S. General Accounting Office (GAO). (2004). Data Mining: Federal Efforts Cover a Wide Range of Uses. GAO-04-548, May 2004. Zhou, Z.-H., Jiang, Y., & Chen, S.-F. (2003). Extracting symbolic rules from trained Neural network ensembles. AI Communications, 16(1), 3-15. Read More