(a) Data Mining Process: In the context of this cluster analysis project, and in your own words, explain how you would execute the first stage of data mining, namely the "Pre-modelling" stage. Be sure to differentiate the sub-tasks in this stage
(b) Pre-modelling: Describe the potential business problem and data mining problem in the context of this project. Be sure to differentiate these two problems in your description.
(c) Data Preparation: Use the "seeds_dataset_twoClass.csv" file to prepare the dataset for cluster analysis. You can use the following table format to justify the data type (i.e., measurement) and direction (i.e., role) used for each attribute.
|
Attribute
|
Data Type
(or Measurement)
|
Direction (or Role)
(Input, Target or None)
|
Justification
|
(d) Data Exploration: Analyse the dataset "seeds_dataset_twoClass.csv" using the following summary statistics in the Data Audit node. Discuss the use of these summary statistics for deciding if further data preparation is required.
a. Mean and Standard Deviation (Std. Dev), Min and Max
b. % Complete and Valid Records
c. Outliers and Extremes
(e) Data Preparation: From the scenario and data given, explain why the attribute A3 (compactness) is probably not useful for cluster analysis. Prepare the data (for mining) by filtering out this field using IBM SPSS Modeller.
(f) Executing Clustering Technique: Decide on the number of clusters (i.e., K) and then execute K-Means on the filtered dataset. Assess the appropriateness of applying K-Means on this dataset. Interpret the clustering results.
(g) Interpreting Clustering Results: Use the Graphboard node to generate a scatter plot based on attributes A4 and A5. The plot should show each data point labelled or coloured based on the cluster number assigned by K-Means. Evaluate the clustering results using this plot (and you may also use the project information given in the Background section of this assignment).
(h) Data Preparation: Having read your preliminary analysis, a colleague gave the following comment: "the dataset should have been normalised before the clustering process." Evaluate the clustering solutions with and without normalisation and then discuss whether normalisation is necessary in this case.