Neural Networks for Business - Paper Example

Neural networks can be described as an intelligent technique which is data-driven and is used to solve very complex, intricate and complicated problems that pertain the identification of patterns. In the last ten years, the use of neural networks has increased especially since they have proved to be useful and, important tools in dealing with a wide range of function areas impacting on businesses. Neural networks have become an important component and, tool in data mining systems. They help in altering the ways in which institutions view the relationship that exists between the business strategy that they have adopted and their data. Neural networks are simple computational tools used in analysis of data and, the subsequent development of models which are used in identifying and, establishing data patterns or structures. These modes are created by data that is referred to as training data. Neural networks are then exposed to these sets of training data, analyze the patterns present in the data sets and can then be used in new data providing a number of outcomes. In addition to that, neural networks, in the context of a business, can be used in:

Is your time best spent reading someone else’s essay? Get a 100% original essay FROM A CERTIFIED WRITER!

Order now

Learning to forecast future events through analysis of the data patterns observed and noted in the historical training data

In the classification of unseen data into pre-defined groups which are based on the features noted in the training data

In clustering the training data into natural groups based on the similarity and likeness of the features in the training data sets.

Association Rule Mining

Association rule mining is a very important concept and, has been the subject of discussion among researchers for a very long time. It is a vital component of data mining. The main function of association rule mining is discovering and, establishing relationships existing in a number of items present in a database. Data mining can be described as a method of extracting information, which was not known before but nonetheless very important and useful, or extraction of patterns from large information repositories, for example, data warehouses and relational database. The main function of data mining is to extract information from a set of data then transform into a data structure which can be easily understood and used in the future. It is a concept that is used in many fields. When incorporated into data mining, association rule mining aims at establishing patterns appearing frequently, interesting correlations and, the associations existing in different sets of items present in transaction databases. Association rules are used in several fields, for example, drug testing, risk and market management, the control of stock and inventory and in telecommunication networks.

Statements which establish the relation and correlation between data present in a particular database are referred to as association rules. Therefore, association rule mining is conducted to establish association rules which satisfy the predefined threshold support and confidence from a particular database. Association rule mining is also used in low cardinal sparse transaction database.

Unsupervised Clustering

Unsupervised clustering can be described as a method of discovering hidden patterns in data. This technique can also be used in establishing intrinsic patterns that are present in data. Unsupervised clustering is used in establishing inferences present in datasets. These datasets have input data which has not have any labeled responses. In addition to that, cluster analysis is also applied in exploratory data analysis for the purposes of establishing hidden patterns in data. In supervised clustering, these clusters are modeled through a measure of similarity whose parameters are probabilistic distances or Euclidean distances. Other applications for unsupervised clustering include; bioinformatics (sequencing analysis), medical imaging and, in data mining. It is also used in detecting gradual change of patterns over time.

PROBLEM 2

De Telegraaf is Dutch Company which published magazines and newspapers. For every issue that has been published, be it a newspaper or magazine, the company has to perform an estimation of the number of issues which will be sold. The company might suffer a loss of investment due to too many issues being distributed without the market. A sell out results in profit decline and, many unsatisfied customers. The best way for De Telegraaf to prevent this to be produce and distribute number of issues which will be sold and, ensure that each and every customer has had a copy (i.e., prevent a shortage). However, it is not possible to produce the exact number of copies needed, only an approximation of the quantity is feasible and, this is because newspaper sales are determined by chance. Currently, De Telegraaf uses multiple linear regression technique.

Neural network application

Neural computing has the ability of providing predictions which are more accurate and precise. What is required is several neural networks, each one representing individual sales point, and they have to be trained with regards to the sales figures of the past five years or so. It is important the training procedure be fully automated. The whole process is designed in a manner which all the unsold and undesirable newspapers and magazines are avoided by the neural network. Through the application of neural networks, De Telegraaf could reduce sell outs by a margin of more than 45% even without increasing the number of issues which remain unsold.

PROBLEM 3

Chest pains (x); Exercises (y); Smokes (z)

K=2

Initial Centroid points = Patient 1 and 3

Yes=1; No=0

Patient ID Chest Pains (x) Exercises (y) Smokes (z)

1 1 1 0

2 1 0 1

3 0 0 1

4 0 1 0

5 1 1 1

6 0 1 1

3X23Y14Z2=3!2!1!3!1!2!4!2!2=112=4

These four sets have to be stored

PROBLEM 4

Patient ID Chest Pains (x) Exercises (y) Smokes (z)

1 1 1 0

2 1 0 1

3 0 0 1

4 0 1 0

5 1 1 1

6 0 1 1

MAVR=2

The numbers of sets were originally 4

42=8

The combinations of the MAVR are (x1,x2),(y1,y3), and (z2,z3).

0: [1,2,3],1: [2,1,2],2: [1,0,2]

References

Zou, H. F., Xia, G. P., Yang, F. T., & Wang, H. Y. (2007) An Investigation And Comparison Of Artificial Neural Network And Time Series Models For Chinese Food Grain Price Forecasting. Neuro computing, 70 (1618), 29132923

Hanafizadeh, P., Poursoltani, H., & Saketi, P. (2007) Comparative Study of Prediction Capability of Artificial Neural Networks Using Early Interruption Method and Autoregressive Time Series Process in Estimating Inflation Rate. Journal of Economic Research, 42(2), 25-36.

Menhaj, M. (2008). Fundamentals of Neural Networks Tehran, Iran: Publications of Amir Kabir Industrial University

Gupta JND, Sexton RS. Comparing backpropagation with a genetic algorithm for neural network training. OMEGA The International Journal of Management Science 1999;27:67984.

Minaei-Bidgoli, B., Tan, P., Punch, W.: Mining interesting contrast rules for a web-based educational system. In: Proc. of the Int. Conf. on Machine Learning Applications (2004) 1-8.

Agrawal R., Imielinski, T., Swami, A.N.: Mining Association Rules between Sets of Items in Large Databases. In: Proc. of SIGMOD (1993) 207-216.

Dougherty, J. Kohavi, M. Sahami, M.: Supervised and unsupervised discretization of continuous features. In: Int. Conf. Machine Learning (1995) 194202.

Hubscher, R., Puntambekar, S., Nye, A.: Domain Specific Interactive Data Mining. Workshop on Data Mining for User Modeling at UM07 (2007) 81-90.

Nitin Gupta, Nitin Mangal, Kamal Tiwari and Pabitra Mitra, Mining Quantitative Association Rules in Protein Sequences Data Mining, LNAI 3755, pp. 273-281, 2006 (c) Springer- Verlag Berlin Heidelberg 2006.

Pratima Gautam and K.R. Pardasani, Algorithm for Efficient Multilevel Association Rule Mining In (IJCSE) International Journal on Computer Science and Engineering, Volume 02, No. 05, 1700-1704, 2010