Global Journal of Computer Science and Technology, C: Software & Data Engineering, Volume 22 Issue 2

, is the number of feature is the number of times the ith feature occurs in the document is the probability of the ith feature occurring given class In text classification, the conditional feature probabilities are estimated using Laplace smoothing. Laplace smoothing can be written as, b) Logistic Regression Logistic regression is a classification model that models a binary variable using a logistic function [31]. It is a categorical dependent variable prediction approach that uses a collection of independent variables to forecast a dependent variable. Fig.1: Example of a Logistic Regression Because of the application of the sigmoid function in logistic regression, the curve in Fig. 1 is generated. The sigmoid curve is another name for the curve produced above. c) Support Vector Machine Vapnik and Cortes proposed a Support vector machine (SVM). The Support vector machine is a supervised machine learning method used for classification, regression problems. It determines the best decision boundary among various vectors [12]. Support Vector Machine draws a hyper-plane to separate two distinct classes [25]. Fig. 2: Support Vector Machine parallel with the hyper-plane and the distance between two marginal lines with the hyper-plane is equal. The total distance (D) between the two margins is called to be ma rginal distance. For having a better model SVM finds the largest distance (D) between support vectors in shorts marginal lines. SVM produce a linear hyper-line of separation between classes which are linearly separable. But in nonlinear classification problems like as Fig. 3, it is not possible to have better classification only by drawing a hyper-plane. To solve Fig. 3a this kind of problem support vector machine using the so-called kernel trick. They have the capability to implant the data from a lower dimension into higher-dimensional space. In Fig. 3b we can see that, the kernel trick takes the lower space input points Fig. 3a and then implant them into a higher dimension. Fig. 3: A Non-linear classification problem. d) Passive Aggressive Classifier The passive-aggressive algorithm is a member of Machine learning algorithms. Passive- Aggressive algorithms are called so because : P L aplace ( f i | y c ) = λ + P n d j =1 x j i P n f k =1 [ λ + P n d j 0 x k j 0 ] p ( x j | y c ) = π n f i =1 p ( f i | y c ) x i i Y X w ∗ x − b = 0 w ∗ x 1 − b = 1 w ∗ x 1 − b = − 1 D (a) Non-linear data point. (b) Solving Non-linear classification problem using Kernel. Fake News Detection: Covid-19 Perspective Global Journal of Computer Science and Technology Volume XXII Issue II Version I 3 Year 2022 ( ) C © 2022 Global Journals n f jth x j p ( f i | y c ) x i i y c Here, is the number of data points in the training set from class , is the number of features , is a parameter known as the Laplace smoothing constant. is typically set to 1. As is a non-zero number, it prohibits such degenerate cases from equaling zero. In general, if a feature does not occur in any document of the training set, or all documents The test set that contains the degenerated feature will be zero for all classes , all discriminating power. λ λ λ p ( x j | y c ) x j y c y c n f n d causing Multinomial bayes to lose Naïve . class given the document that is formed by Bayes rule, Multinomail bayes uses multinomial model to find out . So, in perticular y c x n p ( y c | x j ) = p ( y c ) p ( x j | y c ) p ( x j ) x j p ( f i | y c ) , Naïve Support Vector Machines not only create hyper- plane but also construct a maximum margin separator, a decision boundary, with the largest possible distance. SVM constructs the marginal line Fig. 2 by the nearest one or more positive or negative points, these points are called to be support vectors. The Marginal lines are drawn