Faculty of Business and Law
Assessment Information
This assignment is an individual work and it is designed to assess learning outcomes 1, 2, 3, 4 and 5:
Intended learning outcomes
1. Define and evaluate key concepts of business analytics.
2. Critically apply business analytics skills for decision making.
3. Critically analyse and interpret the outputs of data mining models and forecasting results for end-users.
4. Solve managerial problems and make systematic decisions by applying business data analysis techniques.
5. Have ability to apply business analytics to various international business contexts by selecting appropriate techniques.
Task One-Business analytics concepts (Clustering Analysis) [15 marks]
a. Critically analyse and explain the concepts and applications of Clustering analysis different tools and methods. Some samples from proper resources should be provided. The applicability and advantage of clustering analysis should be described and analysed properly in a real case published in peer reviewed journals. [15 marks]
Task Two-Marketing Analytics [40 marks]
Suppose that you are working in a Telecommunication industry as a business analyst. Your line manager has asked you cluster the companies of mentioned industry. The companies of each cluster should be similar while companies from different clusters should be different. You have to select at least 15 companies and 5 variables. Then try to use the K-means and Hierarchical clustering analysis to cluster selected companies. Compare the results of two methods and discuss on proper number of clusters. After clustering, describe similar characteristics in each cluster and finally select proper names to extracted clusters. In this task you should describe your analysis and interpret the results properly. You can use a software, but any related figure and results should be interpreted properly in your report.
Task Three: Forecasting [45 marks]
You have been asked to forecast the total revenue of companies of the first cluster extracted in the previous task for 2019. Prepare a table containing yearly total revenue of the cluster (2005- 2018) do the following items:
a) Use Exponential Smoothing (with smoothing constants of 0.3, 0.5 and 0.7) to forecast the revenue of the cluster for 2019. Compare the forecasting accuracy of three mentioned constants and select the proper smoothing constant using the MAD.
b) Use Naïve method, Moving average (3) method, Weighted average (W1=0.5, W2=0.35, W3=0.15) method, and Exponential Smoothing method (with the proper smoothing constant determined in part (a)) for forecasting. Compare the forecasting accuracy of mentioned methods by the MAPE.
c) Determine the forecasted total revenue of the cluster for 2019 by Naïve, Moving average, Weighted average, and Exponential Smoothing methods.
d) Determine that which method is more accurate for forecasting the total revenue of the cluster for 2019? How much is your forecasted the total revenue value for 2019?
Answer:
Task One- Business Analytics Concepts
The process of partitioning a set of data objects or observations into subsets is named as cluster analysis or clustering. The objects which are grouped in a cluster are different from the objects present in other clusters but similar to each other and each subset is known as a cluster while the clustering is the set of clusters. Humans are not the ones who carry out this partition; it is done by the clustering algorithm. The discovery of previously unknown groups is also possible within the data that is why it is considered as very useful.
Applications:
Out of various types of application of cluster analysis few are web search, business intelligence, security and pattern recognition of images. In business intelligence, Organization of a large number of customers into groups can be done using clustering where the customers within a particular group are strongly similar to each other. This helps in facilitating the development of business strategies for a good customer relationship management for the future. Clustering can be applied in order to partition projects into category-based n similarities to improve project management which is required for the effective conduct of project auditing. In image recognition, discovery of clusters or subclasses is done by this method in handwritten system which is used for recognition of characters. There is a large variance in the way people write a particular digit. For instance, we can consider a number 2, which is written in different ways by different people, here clustering can be used for determination of subclasses of “2”, each of which will represent a variation in the way in which this digit can be written. Overall recognition accuracy can also be improved by the use of multiple models which are based on the subclasses. In web search also, clustering has found many applications. For instance, due to the availability of a large number of web pages, a single key word search can lead to a result of large number of hits which are the pages relevant to the search. The search results obtained can be organized into groups and be presented in a concussed and conveniently accessible way. For clustering of the documents into relevant topics, these techniques have been developed and are commonly and widely used.
In the field of data mining, for getting insight of the distribution of data, cluster analysis can be used as a standalone for focusing on a particular set of clusters for further analysis. It may also be considered as a preprocessing step for other algorithms like attribute subset selection, characterization and classification which would operate on the detected clusters.
Classification:
As the cluster is a collection of similar data objects grouped in a particular cluster and different data objects in another cluster, a cluster of data objects is more appropriate to be treated as an implicit class. Following this logic, it is sometimes named as automatic classification. The property of automatically finding the grouping is a very rare advantage of cluster analysis. The classifications of different samples of varied subjects can be segregated into different groups by a statistical multivariate method known as clusters in order to analyze is what we call cluster analysis. The classification must be done appropriately while implementing this analytical process to segregate them into different groups they are measured according to the varying set of variables. Hierarchical and non-hierarchical are two methods in which the statistical methods of clustering analysis have been classified. If one of these methods is not applied than it must be noted that performing a clustering analysis is not possible. Agglomerative method is one of the methods in which the hierarchical method has been divided. Respective cluster gradually formed by subject. Between the two most similar clusters the procedure of integration is also involved, to make sure that separate subjects are categorized into their respective order this procedure must be continued. The optimum number of clusters is selected among varied segregated groups at the end of this analysis. One more type of hierarchical cluster analysis is termed as divisive method. A single cluster is formed or collected when the subjects are bound in this type of hierarchical analysis. It is considered that this method is the reverse method of agglomerative method. ‘K- Means’ technique (Cornish 2007) is a technique by which the non- hierarchical method is being mostly recognized. Effectiveness in terms of segmenting different types of customers can be delivered on behalf of the company while applying a clustering analysis concept. Customer subdivision can also be done successfully into small number of groups using this process. Company can examine its performance in various marketing activities through clustering analysis so this method is considered beneficial for the companies. This method favors in delivering effective branding strategies and attaining competitiveness so this concept is highly valued by marketers in business-
to- business activities. Business products are also saved from being unsold by the marketers through these benefits. In the real case highlighted by Mudambi (2002) this can be promptly evident. Three varieties of segmentation were identified by a flat- rolled steel industry of the North America in this study with the help of cluster analysis, and the segmentations were commitments, services and sensitivity among prize.