Regression Analysis
-YASH BHANGARE
-BECOA110
Regression analysis gives the relationship between dependent variable and one
& more independent variable.
Linear regression:
Dependent variable is continuous in
nature.
Generalized model used to predict
categorical variable.
Linear Regression Equation;
Y = mX + C
Linear Regression :----Simple Linear Regression
----Multi-Linear Regression
1 independent variable and 1 dependent variable
More than 1 independent variable and 1 dependent variable
Y = β1 + β2X2 + β3X3 + …… + βkXk + ε
-----where,
Y = dependent variable
X = independent variable
β1 = Regression coefficients
Ε = Error term
β
● When actual means of x & y are given,
● When assumed means of x & y are given,
Logistic Regression:
Dependent variable is binary in nature and independent variable can be
continuous or binary.
When more than two categories are involved in dependent variable - multinominal
logistic regression.
Difference between Linear regression & Logistic regression
Sr.no. Linear Regression Logistic Regression
1 y= a + bX y = e^(b0 + b1*x) / (1 + e^(b0 + b1*x))
2 Works with single independent
variable
Works with more than one variable
3 Estimated using OLS Estimated using MLS
4 Continuous output Constant output
5 Predicting housewife or stock price Predicting whether a patient has disease
or not
Overfitting occurs when algorithm captures the noise of data.
Overfitting - algo works well on training set and fails on testing dataset.
Underfitting occurs when algorithm cannot captures the underlying trend of data.
Underfitting - algo failed to work on both training and testing dataset.
Regression Models
❏ Polynomial regression
❏ Ridge regression
❏ Lasso regression
❏ Elasticnet Regression
Thank You

Regression-types

  • 1.
  • 2.
    Regression analysis givesthe relationship between dependent variable and one & more independent variable. Linear regression: Dependent variable is continuous in nature. Generalized model used to predict categorical variable. Linear Regression Equation; Y = mX + C
  • 3.
    Linear Regression :----SimpleLinear Regression ----Multi-Linear Regression 1 independent variable and 1 dependent variable More than 1 independent variable and 1 dependent variable Y = β1 + β2X2 + β3X3 + …… + βkXk + ε -----where, Y = dependent variable X = independent variable β1 = Regression coefficients Ε = Error term β
  • 4.
    ● When actualmeans of x & y are given, ● When assumed means of x & y are given,
  • 5.
    Logistic Regression: Dependent variableis binary in nature and independent variable can be continuous or binary. When more than two categories are involved in dependent variable - multinominal logistic regression.
  • 6.
    Difference between Linearregression & Logistic regression Sr.no. Linear Regression Logistic Regression 1 y= a + bX y = e^(b0 + b1*x) / (1 + e^(b0 + b1*x)) 2 Works with single independent variable Works with more than one variable 3 Estimated using OLS Estimated using MLS 4 Continuous output Constant output 5 Predicting housewife or stock price Predicting whether a patient has disease or not
  • 7.
    Overfitting occurs whenalgorithm captures the noise of data. Overfitting - algo works well on training set and fails on testing dataset. Underfitting occurs when algorithm cannot captures the underlying trend of data. Underfitting - algo failed to work on both training and testing dataset.
  • 8.
    Regression Models ❏ Polynomialregression ❏ Ridge regression ❏ Lasso regression ❏ Elasticnet Regression
  • 9.