Decision Support Systems - Artificial Neural Network - Essay Example

Dесisiоn Suрроrt Systems Name Institution Artificial Neural Network (ANN) ANN is a model that processes information based on the principle of biological nervous system. The relationship between the two processes is their functionalities. The model is based on various elements representing the neurons and work in unison to help solve specific problems. Just like the human beings, ANNs learn through examples. ANN is configured in such a manner that it is compatible with specific applications since it has the ability to recognize a given pattern or classify the data in a given manner. In relation to biological applications, the neurons functions properly through adjusting its systems to the synaptic connections existing between the neurons. Therefore, the model is designed in a way that it constitutes flexible nonlinear models of mimicking biological neural system. Initially, ANN did not receive much attention due to limited applicability. Moreover, its capacity of computing at a time was limited. Through permitting simple functions to rise indefinitely, the multi-layered ANN has the ability of approximating a large class of functions to any desired extent of accuracy. Organizations use ANN in different ways including signal processing, language leaning, and pattern recognition. Since human often solve problems based on experiences, ANN works in similar manner by identifying previously solved problems and looking into the pattern to assist in finding the right solution. Moreover, ANN has the ability to learn the patterns and correctly classify them. On the other hand, neural network is also capable of assembling human characteristics in problem solving and assimilate them logically using analytical techniques, and standardized software technologies. ANN uses a number of inputs, weigh and sum them up then use the outcome for arguing the singular-valued function from the neurons output (Metz, 2015). The model uses different approach in solving the problem compared to conventional computers, which uses algorithmic approach. Algorithmic approach involves the use of a given set of instructions. It is important to note that neuron networks cannot be programmed to perform a specified task; hence, it is important to carefully select the examples to avoid time wastage. Generally, ANN is computer-based models. This modelled brain bring on board simple methods of solving technical problems. Even though computers play major role in modern societies, they have troubled recognizing simple patterns. ANN constitutes of simple processing units wired together within a complex communication network where every unit represents a simplified form of model. These units represent the neurons in biological network and send off the signals whenever they receive enough input signal from the other interconnected units. Additionally, the units are structured in a manner that they form interconnected layers for solving the problems. Even though some networks only have one layer or element, in order to function properly, most applications required at least three layers constituting of the input, hidden, and output. The input layer of neurons receives the data from the input files or from the electronic sources, which in most cases are sensors within real time applications. Upon receiving the information, the output layer then sends it directly a secondary computer, which is considered to be in the outside world. Besides a secondary computer, the information might as well be channeled to other devises functioning as mechanical control centers. Between the input and output, layers there could be several hidden layers, which in most cases are the internal layers. These hidden layers have neurons within their interconnected structures. In many networks, the hidden layers often receive signals originating from all neurons in the layers above which mostly are the input layers (Metz, 2015). After performing all its functions, the neuron then passes all the outputs to a layer below it thereby providing a feed forward path to the output. The manner of connection of the neurons plays a crucial role on the impact of operations within the network. Figure 1: Interaction of ANN components Figure 2: Output generation process of AAN Comparison and contrasting of Artificial Neural Networks (AAN), Support Vector Machines (SVM), and k-Nearest Neighbor (kNN) All the three models play a significant role in comparison and classification of different data. Moreover, the models are based on learning processes through application of different data sets. SVM has the ability of utilizing data from remote sensing and classify the information in an effective manner than kNN and ANN. Moreover, SVM has higher capacity of generalizing relatively smaller numbers compared to other models. SVM is more accurate compare to ANN and kNN in analyzing information relating to maximum likelihood algorithm. Nearest neighbor algorithm is quite simple and uses generalized data based on logic reasoning. Additionally, the model also required little experience of use compared to others, which are quite technical. Since the model is quite simple, it is widely used in real world applications and act as the basic component of the complex models. SVM is mainly used for binary classification. All the models have the ability to realize pattern recognition occurring between the layers through finding a decision function. Within the model, this is possible by selecting the points from the data. Both kNN and ANN mainly focus on minimizing the risks associated empirical data while SVM focuses on minimizing structural risks based on probability functions. There are factors influencing the performance of kNN, which include the distance function for determining the nearest neighbors, the decision rules used in deriving the function, and the number of neighbors used in the classification of new example. The major weakness associated with SVM is that the result from classification is purely dichotomous. Moreover, the model does not give any probability to the data. On the other hand, kNN weakness is the neighbor has the ability of providing explanation for the result of classification. However, this might be a strength in areas where black box models are considered inadequate. Another drawback associated with the model is in the calculation the neighborhood where there is need to define a metric that measures distances between data items. Majorly kNN focuses on trials and errors while defining the metrics associated with different data, which might result in inaccurate information. Another strength of the model is that it uses simple information, which could be manipulated to yield adaptive behaviours. Moreover, the model also lends itself parallel to implementations. Nevertheless, kNN requires a wider storage when used as a classifier. SVM on the other hand is complex and requires an extensive memory making the testing speed quite slow. The strength of ANN is that it has higher ability of tolerating noisy data. Additionally, it has several training algorithms. However, the model has greater computational burden. Besides, it is susceptible to over fitting and requires longer duration of training. Application of ANN in businesses Every company needs to increase its value in order to create more opportunities that are sustainable and would ensure positive organizational performance. However, it is important to consider both external and internal factors while conducting valuation process of the company. Besides those factors, it is important to have knowledge on the methods of driver value forecasting. As a result, most organizations are focusing on ANN as a method of simulating their data and. It is possible to apply the model and solve the problems within any business environment conventionally. Additionally, the model has helped solve problems faster and better without human interventions. To some extent, the model has made it easier for companies to arrive at the solutions of complex situations. Applications of the model vary from financial institutions, insurance companies to oil drilling companies. Most financial institutions use the model to improve their decisions through improving interpretational of behavioral scoring systems and developing models relating to credit card risk and bankruptcy management. Companies that have a focused on developing the model are the international airports like New York, London, and Miami. These measures are in place mainly to increase the levels of security especially with the rising levels of terror threats. Before boarding the planes, these airports export the bags to undergo the unusual thorough inspection before being loaded into the cargo. Besides using the x-ray and metal detectors for detecting metallic weapons, these airports have employed the use of ANNs in screening the plastic explosives. Moreover, this detection system bombards the luggage with the neurons and displays the emitted gamma rays in response. After the displays, the network the analyses the signal and decide whether the reported response predicts an explosive. The main aim of using such detective measures is to help detect the explosives with a 95% probability while at the same time minimizing the number of false alarms. It is important to note that detection of the explosives using the gamma rays is very difficult considering the fact that numerous chemicals elements are released at different frequencies. In order to minimize the classification error, these airports have supervised trainings. As a result, the trained networks used with the ANN model has been able detect security threats. However, it is crucial to note that reduction in the number of false alarms reflects the fewer number of bags that should be opened and examined. With all these factors in place, the airports have been able to reduce their operational cost, improved efficiency within the check-in process, and customer satisfaction that is the aim of every business. Reference Metz, C. (2015, April 21). Finally, Neural Networks That Actually Work | WIRED. Retrieved from http://www.wired.com/2015/04/jeff-dean/ Read More