The Best Accuracy and Speed of the WEKA Software - Assignment Example

Business Intelligence Analysis The present day car-sales business is entirely composed of raw and unorganised data that is to be translated into meaningful information for the business survival. Proper, effective and efficient use of this business information results translates to competitive advantage by enhancing the decision making process and optimising business processes that help in the prediction of market dynamics and optimised forecast of resources. Business intelligence (BI) study focusses on detailed study of data mining techniques by putting more emphasis on classification techniques as the most vital supervised learning models. WEKA is the preferred software tool for the classification. This is because the software can mine useful information from the data by providing an algorithm for coming up with accurate prediction models. Business Intelligence Analysis Introduction Overview of the course The objective of the business intelligence analysis is to equip learners with business skills that will help them in relating services, applications, and the technologies so that they can manage and analyse data so that the data is transformed into useful information that is critical to sound decision making. The Business intelligence processes are used to evaluate business processes and to optimise them so that they are in line with the objectives of the business. The objectives can be operational that impact the business on daily basis, or even tactical that will have impacts on short term basis like marketing or even strategic goals that will have long term effect like improving the revenues. Business intelligence is therefore a critical element in the decision making process. This is because the process analyses data that is, orders, inventory and points of sales. From these data the Key Performance Indicators (KPI) are calculated that is, the measure of the strategic performance of the business. KPI is important in the monitoring of the organisation’s performance and for decision making. The main objective is to extract business intelligence that is crucial in shaping the organisation’s operations to the strategic and tactical objectives of the business. This is only possible when the transactional data is integrated in the data warehouse to be analysed by the BI performance managers. The course seeks to equip learners with the ability to convert the integrated, but raw warehouse data into useful information that can be used to determine accurately how the business performed in contrast to its objectives and predictions. This application generates business knowledge that is very important for planning. Technology has enhanced modern day car sales business at the cost of expanding the information required for selling. This has necessitated the creation of databases to store the information. The creation of knowledge and management (data mining) has become a critical element in the car sales business. Data mining is the activity of finding valid and crucial patterns within given data sets? Data mining is often classified as either supervised or unsupervised learning. The supervised learning technique requires the building of a model that is used in the analysis such as in classification, regression and association rules. Unsupervised learning is not guided by variables such as clustering. In car sales business emphasis will be on classification. It is the most widely used technique for data mining technique. The main objective will be to predict the target class for each case in the data. Methodology Comparisons of the classification techniques Decision trees They are an easy to use but very powerful classification algorithms. They are used to recursively divide the information by the use of depth first method up to the stage where the information has been divided according to their classes. Their advantages is that they are easy to use and can handle both serial and parallel data. The only disadvantage is that there is a limitation on the number of the output attribute to one. Bayesian network They are graphical models that are used to simplify the computations involved in car sales. Their advantage is that they exhibit a high accuracy. Simulations The data was extracted from car sales information. The data warehouse is comprised of elementary data of all the items sold, their size and number of items to be sold. The attributes are as shown below Attribute lists: Buying: vhigh, high, med, low. Maint: vhigh, high, med, low. Doors: from 2, 3 4, 5more. Persons: from 2, 3, 4, more. Lug_boot: small, med, big. Safety: low, med, high. classValues: unacc, acc, good, vgood. 1.2 Identify your input and class variables (which variable are you going to use as your class variable) There are 7 attributes and a total number of 1729 instances. The “classValues” attribute will be used as the class variable. Comparison of data mining tools There are varied tools available in the market such as Rapid Miner, Knime, Tanagara, Orange and WEKA. Of all the tools, the chosen tool, that is WEKA, has the best performance in accuracy. Determinately, WEKA, has the capacity to deal with multi class data set. This formed the basis of the choice because the other software’s could not handle multi class data sets. WEKA, also beats the other data mining tools because of its applicability and its ability to run classifiers with all the data sets. WEKA data mining software The WEKA data mining tool consists of many machine learning algorithms for data mining. It can be used for data pre-processing, regression, classification, classifying, feature selection and visualisation. WEKA interfaces WEKA consists of four interfaces as shown below. The explorer is the main interface that is used to run simulations. Files of the format ARF, CSV, C4.5, and library are loaded through the exploerer. Loading the car sales .csv file produces the above output. Obviuosly there are 6 tabs for preprocessing, classifying, associating etc. The preprocessing tools are used for filtering data. Data is obtained from files and the pre-processor is used to draw histograms of the attribute in question. The data can also be normalised in the explorer. The classify tools are employed to continue with the analysis of the data. For data that requires classification or regression, it is analysed by the classify tab. WEKA for car sales analysis Car sales is the field of research that is concerned with the study of organising the information associated with car sales. Typically, the information can be depicted as machine learning tasks that is, classification, regression etc. WEKA tool provides algorithms for such types of problems. Iinterestingly, the freeware gives satisfactory coverage and confidence. The software can also be used for determining the for classifying markets and customers data. Because the WEKA data mining tool offers the information that is needed for deeper understanding of market problems and allows the business to solve the problems. The tool was used to build decision tree models to work on the customer evaluation. The tools produced precise but very efficient descriptive models of the car sales files. Results from the analysis Analysis was carried out to compare the classification techniques on the different .csv files. The car sales.csv file was first loaded. From the results it is very easy to distinguish between the constraints. Weka divides the simulation into several subparts so that the analysis process is simplified. The first sub item contains the accurately and the inaccurately classifed instances stored in numeric and as percentages. The data set was analysed with ten folds cross validation. WEKA computed all the needed parameters with respect to the classifiers accuracy and prediction rate. Evidently, the Bayesian classifier had the highest accuracy that is, 98.6111% with a time of 0.17 secondswhile the Decision tree had the least that is, 81.9444% with 0.62 seconds. Deducing from this data, the Bayesian is clearly the best strategy among the three because of its hiogh accuracy and its fast capability. The car data set has 1729 instances and 7 attributes. === Run information === Scheme:weka.classifiers.rules.ZeroR Relation: CarDataSetExcel Instances: 1729 Attributes: 7 buying maint doors persons lug_boot safety classValues Test mode:evaluate on training data === Classifier model (full training set) === ZeroR predicts class value: unacc Time taken to build model: 0.01 seconds === Evaluation on training set === === Summary === Correctly Classified Instances 1210 70.0231 % Incorrectly Classified Instances 518 29.9769 % Kappa statistic 0 Mean absolute error 0.229 Root mean squared error 0.3381 Relative absolute error 100 % Root relative squared error 100 % Total Number of Instances 1728 Ignored Class Unknown Instances 1 === Detailed Accuracy By Class === TP Rate FP Rate Precision Recall F-Measure ROC Area Class 1 1 0.7 1 0.824 0.5 unacc 0 0 0 0 0 0.5 acc 0 0 0 0 0 0.5 vgood 0 0 0 0 0 0.5 good Weighted Avg. 0.7 0.7 0.49 0.7 0.577 0.5 === Confusion Matrix === a b c d Read More