Artificial Intelligence for Big Data
上QQ阅读APP看书,第一时间看更新

Generalized linear model

While we have tried to understand the concept of linear regression with one dependent and one independent variable, in the real world, we are always going to have multiple dependent variables that affect the output variable, termed multiple regression. In that case, our y = a + bx linear equation is going to take the following form:

y = a+ b1x1 + b2x2 + ...+ bkxk

Once again, a0 is the y intercept, x1, x2, ...xk are the independent variables or factors, and b1, b2,.., bk are the weights of the variables. They define how much the effect of a particular variable has on the outcome. With multiple regression, we can create a model for predicting a single dependent variable. This limitation is overcome by the generalized linear model. It deals with multiple dependent/response variables, along with the correlation within the predictor variables.