edobander（7页）.docVIP

Edo Bander Math 4500 Research Project Bayesian Algorithm Executive Summary Bayesian learning algorithms are methods of calculating probabilities for hypotheses and are one of the most practical approaches to certain type of learning problems. Bayesian classifier is competitive with other learning algorithms in many cases and in some situations, it outperforms other methods. The reason why Bayesian algorithm is important to machine learning is because it provides a perspective for understanding other learning algorithms. Bayesian analysis is also used to justify a design choice in neural network learning algorithms. This research analyzes a CardiologyCategorical.xls data set. This data set contains 302 different instances of 14 different attribute. These attributes are: age, sex, chest pain type, blood pressure, cholesterol, blood sugar, resting heart rate, angina, peak, slope, colored vessels, thal and class. The contribution of all attributes is independent and each contributes equally to the classification problem. By analyzing the contribution of each attribute, a conditional probability is determined. The classification that will be analyzed is made by combining the impact that different attributes have on prediction. Given the data set, first we estimate the prior probability for each class by counting how many times each class appears in our data set. For each attribute, a, the number of occupancies of each value a can be counted to determines P (a). When we classify certain instance, the conditional and prior probability generated from data set is used. This is done by combining the different attribute values from the given instance. This would be known as a conditional probability or P (T | a). An example that we’re going to use for instance is T = [ M, 154 ] from our data set. This means that we’re trying to classify in this situation a chest pain type from attributes Gender (male) and Heart rate. At this point, we have both prior and con