Classifier Model - Real-World Data - Attribute Selection Method - Test Dataset - Assessment Answer

Assessment Task:

The goal of this project assignment is to give students an opportunity to build and test a classifier model using real-world data. The data to be used was downloaded from UCI Machine Learning Lab and it was slightly modified by removing some attributes and discretizing the class attribute. 
You may use any tool or software for preprocessing but you must use Weka to select attributes, build models, and test models. 
You are given three files: project-crime-binary-reduced.arff, crime-data-nv-all.pdf, and save-and-load-model.pdf. 
The project-crime-binary-reduced.arff file has the dataset for the project and it has 1819 instances and 89 attributes. 
The crime-data-nv-all.pdf file has the description of the dataset. Note that this file has a description of all attributes in the initial dataset. The dataset given to you does not have all these attributes. 
The save-and-load-model.pdf file describes how to save a classifier model and load it later to apply to a test dataset. 

Requirements: 
Among 89 attributes, the first two attributes are not used to build a model and the last attribute is the class attribute. So, you first need to remove the first two attributes and use the remaining 87 attributes to build models. 
Your goal is to select between 5 and 15 predictive attributes from the 86 non-class attributes, and a classifier algorithm that would achieve a high classification performance. Follow the instructions given below. 
1. Randomly select about 600 instances and set them aside as a test dataset. When you sample the test dataset, you need to make sure that the class distribution in the test dataset is approximately the same as that in the initial dataset. This means that you must perform a stratified random sampling. The remaining about 1219 instances will be used as a training dataset. Note that the training dataset and the test dataset must be independent of each other (i.e., there should not be overlap between the two). In step 3, you will modify these datasets by removing some attributes from them. Since you have to repeat the whole process multiple times, you have to reuse this training dataset and test dataset. So, you may want to keep one copy each of unmodified training dataset and test dataset. 
2. Select between 5 and 15 predictive attributes from the training dataset using an attribute selection method that is implemented on Weka. The reason I specified 
“between 5 and 15 attributes” instead of a fixed number of attributes is that different attribute selection methods may give you the different number of attributes and some attribute selection methods may give you only the ranks of the attributes (instead of selecting some attributes for you). 
3. Once attributes are selected, remove all other attributes from the training dataset except the selected predictive attributes and the class attributes. If the number of attributes selected by an attribute selection method is n, then now you have n + 1 attributes. For example, suppose that an attribute selection method gives you 7 attributes. Then, you will have 8 attributes (7 selected attributes plus the class attribute) after removing all other attributes. You also need to remove the same attributes from the test dataset. Note that both training and test datasets must be arff files and the format of the test dataset must be exactly the same as that of the training dataset. What this means is that the number of attributes and the names of the attributes must be identical, the order of attributes in both arff files must be identical, and the order of attribute values of categorical (or nominal) attributes must be identical. It is also very important that the format of your datasets must be exactly the same as the format of the dataset, project-crime-binary-reduced.arff, except that your datasets have a smaller number of attributes. This is because an independent test dataset which will be used to test models have exactly the same format as that of the original dataset. If your datasets have different format, then I cannot test your models. 
4. Build a classifier model using the training dataset. You need to try at least following 
ten classifier algorithms: J48, NaiveBayes, IBk (this is KNN; run with k = 10), Bagging with J48, Bagging with NaiveBayes, Bagging with IBk (again run with k = 10), AdaBoostM1 with J48, AdaBoostM1 with NaiveBayes, AdaBoostM1 with IBk (again run with k = 10), and MultilayerPerceptron (neural network). Note that building a model using the MulilayerPerceptron algorithm will take a little longer. 
5. Test your model on the test dataset which has 600 instances (refer to the save-and- 
load-model.docx file). 
6. Compare the test results from the ten classifier algorithms and choose the one which gives you the highest performance. 
Repeat the whole process (step 2 through step 6) at least two more times with different attribute selection methods. This means you must try at least three different attribute selection methods. So, you have to build and test a classifier model at least 30 times (3 attribute selection methods x ten classifier algorithms). 
Then, choose the model that gives you the best performance. Keep this model along with the training dataset from which the model was built and the test dataset on which the model was tested. 

This IT and Computer Science Assessment has been solved by our IT  and Computer Science  Experts at My Uni Paper. Our Assignment Writing Experts are efficient to provide a fresh solution to this question. We are serving more than 10000+ Students in Australia, UK & US by helping them to score HD in their academics. Our experts are well trained to follow all marking rubrics & referencing style.

Be it a used or new solution, the quality of the work submitted by our assignment experts remains unhampered. You may continue to expect the same or even better quality with the used and new assignment solution files respectively. There’s one thing to be noticed that you could choose one between the two and acquire an HD either way. You could choose a new assignment solution file to get yourself an exclusive, plagiarism (with free Turnitin file), expert quality assignment or order an old solution file that was considered worthy of the highest distinction.