# AnalyticBridge

A Data Science Central Community

Sharath Dandamudi
• Hong Kong

## Sharath Dandamudi's Discussions

### Reasons behind deterioration of r squared for an OLS model over a period of time

Started Nov 16, 2011

Hi!,I am working on a project which involves investigation of deterioration in R squared over a period of time for an OLS model. I am wondering how to approach this problem considering the fact that…Continue

### Computation of Weight of Evidence when either the number of bads or goods in a class of a variable is 0

Started this discussion. Last reply by kiran chapidi Dec 11, 2012.

Hi All!I want to understand the ways in which Weight of Evidence (WoE) is computed or adjusted in the following scenarios: 1. When number of goods in a class of a variable is 02. When number of bads…Continue

### Pros and cons of Dummy variable vs WoE approach for variables in Model building

Started this discussion. Last reply by Sandeep Sunkara Mar 24, 2012.

Hi! I need inputs on the pros and cons of building a log-reg model using dummy variables instead of the Weight of evidence approach for categorical variables. Some of the cons that I can think of…Continue

### Books or Reference material for FRAUD & RISK ANALYTICS

Started Apr 5, 2011

Hi!, Can anyone suggest titles of books or reference material found on the web for FRAUD & RISK ANALYTICS? Regards,SharathContinue

# Sharath Dandamudi's Page

## Latest Activity

"They are two different concepts. Prediction uses explanatory variables to characterize expected outcome or expected response. On the other hand, forecasting uses trends in observation to characterize expected outcome or expected response. In…"
May 31, 2017
"Prediction is the generalize term & it's independent of time. Forecasting is the prediction with time as a one of the dependent variable. Eg-  Prediction- Predicting amount spend by user for certain case. It's happen over the…"
May 31, 2016
"Imagine i ask u to choose 5 numbers that sum up to 100. For simplicity sake u tell me the 5 no.s are 20, 20, 20, 20, 20. But when you utter 20 fourth time, i tell you to stop and ask you what was our goal you tell me that the goal was to choose 5n.o…"
Jan 23, 2016
"The 'linearity' assumption in linear regression means that the expected value of the response is a linear function of the parameters.   "Linear in the betas."  Compared to "linear in the predictor…"
Sep 15, 2015
"I have tried to use the WOE = ln(bad_distribution/good_distribution) when the age variable age band bads goods 19-25 2388 2019 8 26-30 1920 1716 24 31-35 1399 1377 53 36-40 1097 1157 73 41-45 934 1126 113 46-50 628 948 180 >50 527 876 209 The…"
Dec 11, 2012
"Hi Sharath, Since sensitivity and (1- specificity) are determined using different cut off points from the confusion matrix, you can get both of them by varying the cutoff points..Let's say I have scores of 1 million observations in my model. If…"
Mar 7, 2012
Sharath Dandamudi posted a discussion

### Reasons behind deterioration of r squared for an OLS model over a period of time

Hi!,I am working on a project which involves investigation of deterioration in R squared over a period of time for an OLS model. I am wondering how to approach this problem considering the fact that there are few dummy independent variables as well along with a few continuous ones. Do I have to check if there is any shift at an overall model level and characteristic level? If there are any other approaches that are adopted in the industry then please let me know. Any help on this would be…See More
Nov 16, 2011
"Exactly as you've written - it's undefined for some categories. Such categories can't be used by logistic regression as well. You have several options: - discard attributes having such categories - merge categories so none of them…"
Oct 18, 2011
Sharath Dandamudi posted a discussion

### Computation of Weight of Evidence when either the number of bads or goods in a class of a variable is 0

Hi All!I want to understand the ways in which Weight of Evidence (WoE) is computed or adjusted in the following scenarios: 1. When number of goods in a class of a variable is 02. When number of bads in a class of a variable is 0 WoE = ln(distribution of goods/distributions of bads) Scenario 1: WoE=ln(0) ?? when number of goods in a class =0.Scenario 2: WoE=ln(distribution of goods/0)=ln(infinity) ?? when number of bads in a class = 0.  Regards,Sharath See More
Oct 17, 2011
"I understand well. It's about terminology. WoE=Weight of Evidence is metrics and has own formula. Dummy variable is binary flag created from categorical variable with more than 2 categories.    And again - simpler model is better. If…"
Oct 11, 2011
"Hi Jozo, Thanks for the reply. What I meant by WoE vs Dummy is say for eg. there is a categorical variable (independent variable, of course) with 4 levels- Colour-  Blue, Green, Red and White The two ways I mentioned about is computing WoE for…"
Oct 11, 2011
"First - you can compute WoE for both dummy and categorical variable, they aren't competitors. Second - dummies lower degrees of freedom, produce simpler models and simpler is better according Occam's razor. And maybe also less sensitive to…"
Oct 10, 2011
Sharath Dandamudi's discussion was featured

### Pros and cons of Dummy variable vs WoE approach for variables in Model building

Hi! I need inputs on the pros and cons of building a log-reg model using dummy variables instead of the Weight of evidence approach for categorical variables. Some of the cons that I can think of using Dummy variable approach are: 1. Overfitting2. Interpretation of output I know one of the things that needs to be looked at is the number of unique levels within a categorical variable. But, making reasonable assumptions, in a generic sense I would like to know if there are any pros and few other…See More
Oct 5, 2011
Sharath Dandamudi posted a discussion

### Pros and cons of Dummy variable vs WoE approach for variables in Model buildning

Hi! I need inputs on the pros and cons of building a log-reg model using dummy variables instead of the Weight of evidence approach for categorical variables. Some of the cons that I can think of using Dummy variable approach are: 1. Overfitting2. Interpretation of output I know one of the things that needs to be looked at is the number of unique levels within a categorical variable. But, making reasonable assumptions, in a generic sense I would like to know if there are any pros and few other…See More
Oct 5, 2011
"HI Sharath:   I'm president of Strategic Affairs Forecasting LLC and am a futurist that has made very careful distinctions between prediction and forecast for many years. Here are the key elements in my opinion:   -For a number of…"
Apr 25, 2011

## Profile Information

Short Bio:
Data Mining analyst
Field of Expertise:
Predictive Modeling, Data Mining, Statistical Programming
Years of Experience in Analytical Role:
4
Professional Status:
Technical
Interests:
Networking, Other