PSI - Issue 56

B. Kalita et al. / Procedia Structural Integrity 56 (2024) 105–110 Author name / Structural Integrity Procedia 00 (2019) 000–000

107

3

2.2. Machine Learning Techniques With a series of data points {x(1),..., x(m)} linked to a set of outcomes {y(1),..., y(m)}, we want to build a classifier that can learn to predict y. Prediction type - The following table provides a summary of the various categories of predictive models:

Table 1. Categories of predictive models [4, 5].

An example of a column heading

Column A ( t )

Column B ( t )

Goal

Directly estimate P ( y | x ) Decision boundary

Estimate P ( x | y ) to deduce P ( y | x ) Probability distributions of the data

What’s learned

Illustration

Examples

Regressions, SVMs

GDA, Naive Bayes

2.3. Hypothesis The model we select is indicated by the letter h. The model's output for a given input value x(i) is h(x(i)). A loss function is a function that determines how different two inputs—the anticipated value z that corresponds to the actual data value y—are from one another. An estimator's mean squared error in statistics calculates the average squared difference between the estimated values and the actual value. The expected value of the squared error loss correlates with the risk function (loss) known as MSE. Additionally, MSE is usually always just ever positive (and not zero) due to chance or because the estimator ignored information that might have allowed for a more accurate estimate. Additionally, MSE is usually always just ever positive (and not zero) due to chance or because the estimator ignored information that might have allowed for a more accurate estimate. Machine learning refers to the typical loss on a collection of observed data as MSE as an estimate of the true MSE. Mean absolute error is a statistician's measure of errors between identically matched observations. Comparing expected data to seen data is an example of a Y vs X comparison, subsequent time to initial time, and one measuring technique to another. Thus, the absolute errors | e I | = | y I x I |, where y(i) is the forecast and x(i) is the true value, are averaged using math. A ROC curve is produced by comparing the true positive rate (TPR) and false positive rate (FPR). The percentage of observations among all positive observations that were correctly anticipated to be positive is known as the true positive rate (TP/(TP + FN)). The false positive rate (FP/(TN + FP)) is the proportion of observations that are incorrectly projected to be positive among all negative observations. For instance, the true positive rate in a medical test is the percentage of patients who are actually diagnosed as having the disease under consideration. 2.4. Classification and Regression It is a process of understanding and recognising ideas and objects, then classifying them into specified categories, frequently referred to as "sub-populations." Utilizing these pre-categorized training datasets, ML programs employ a range of methods to organize incoming information into important and acceptable classifications. Machine learning classifiers estimate the likelihood or probability that the data that follows will fall into one of the predetermined categories using the incoming training data. Examining a link between independent variables (features) and a dependent variable is done through regression (outcome). It is a machine learning-based approach to predictive modelling where an algorithm is employed to forecast continuous outcomes. A very popular use of machine learning models, when it comes to supervised machine learning, is to solve regression problems. The link between independent inputs and results is taught via algorithms (dependent variable). After that, the algorithm/model is applied to predict the results of novel, previously unseen input features or to fill in missing data. For this method of training models, the features and output must be labelled data. A crucial