The purpose of this assignment is to perform k-nearest neighbor classification, interpret the results, and analyze whether or not the information generated can be used to address a specific business problem.
For this assignment, you will use the “Adult Incomes” data set, provided in the topic resources.
ABC Survey Company collects data via surveys that it then sells to marketing departments. Marketing departments typically do not like missing data. Since survey takers typically do not like to answer questions regarding their salary, the one question usually missing from the survey results is, “Is your annual salary $50,000 or more?”
You are the analyst who has been tasked with finding a way to impute (i.e., fill-in) the answer to the question, “Is your annual salary $50,000 or more?” This information can best be imputed based upon how individuals answer other survey questions related to their marital status, educational level, occupation, and familial relationship status. If this important question can be accurately imputed, then the worth of the survey data provided by ABC Survey Company increases dramatically.
Using only Marital_Status Education Occupation and Relationship variables, find the number of neighbors (k) that minimizes the error rate. Use a range of k between 3 and 10. Include the k Selection Error Log output when submitting the answer.Using the same variables and the k selected in Question 1, rerun the nearest neighbor model using the feature selection option in the IBM SPSS Modeler. What is the set of variables that minimize the error rate? Include the “Predictor Selection Error Log” output when submitting the answer.Using the value of k and the set of variables that minimizes the error rate, rerun the k-nearest neighbor model. What is the classification table? Include the pivot table output when submitting the answer.Consider the following individual: Marital_Status=Never-married, Education=Masters, Occupation=Sales, and Relationship=Not-in-family. Based on the k-nearest neighbor model from Question 3, how would this individual be classified? Provide the predicted income level (>50K or “<=50K) and explain the process that you used to determine the income level. Include the table illustrating the data when submitting the answer.Describe the model building process you used to determine whether or not a particular survey taker earned an annual salary of $50,000 or more. Include discussion of the accuracy of the k-nearest neighbor model and how it can be used in practice to impute the answer to the question Is your annual salary $50,000 or more?
Predicted PartitionObserved>50KPercent Correct=50k >50K16722442.7%Training82102792.6%=50k Overall Percent16.6%83.4%79.6%
Predicted PartitionObserved>50KPercent Correct=50k >50K20418752.2%Training97101291.3%=50k Overall Percent20.1%79.9%81.1%
Predicted PartitionObserved>50KPercent Correct=50k >50K28810373.7%Training26684376.0%=50k Overall Percent36.9%63.1%75.4%