This project is aimed at providing some sort of model that takes factors like salary, work-life balance, commute time, etc to predict whether or not an employee is likely to quit a company. The data set was created by IBM, as an actual data set would be a violation of privacy rights.
Strengths :
1) Almost all of your data is already transformed into a number, or at the very least, an ordered enumeration. This will help tremendously when trying to actually input your data into a model.
2) This solution holds an immediate business value for companies. There is no doubt minimizing employee attrition will improve company profits.
3) Because the data is simulated, you do not need to worry about "incorrect" data. There is probably a good spread of data, as well.
Areas for Improvement :
1) I could also see this project going the other direction, and employers using it to determine employees who are likely to leave and then, rather than focusing on keeping them, find a reason to fire them, as this route usually ends up being much cheaper for the company.
2) I wonder if having a "universal model" is better than having a model for a specific company. It seems like between companies, there is way too much difference for a model general enough to fit all companies would actually be accurate enough for nay one particular company to make decisions based off of it.
3) I think there are a lot of factors left out of the data set. For example, many times when someone leaves the company, it has very little to do with the company itself but rather with their personal life. Maybe the employee just had a child or their spouse got a job in another city. Because these aren't really represented in your data set, I think the model may end up not being totally accurate.
This project is aimed at providing some sort of model that takes factors like salary, work-life balance, commute time, etc to predict whether or not an employee is likely to quit a company. The data set was created by IBM, as an actual data set would be a violation of privacy rights.
Strengths : 1) Almost all of your data is already transformed into a number, or at the very least, an ordered enumeration. This will help tremendously when trying to actually input your data into a model.
2) This solution holds an immediate business value for companies. There is no doubt minimizing employee attrition will improve company profits.
3) Because the data is simulated, you do not need to worry about "incorrect" data. There is probably a good spread of data, as well.
Areas for Improvement : 1) I could also see this project going the other direction, and employers using it to determine employees who are likely to leave and then, rather than focusing on keeping them, find a reason to fire them, as this route usually ends up being much cheaper for the company.
2) I wonder if having a "universal model" is better than having a model for a specific company. It seems like between companies, there is way too much difference for a model general enough to fit all companies would actually be accurate enough for nay one particular company to make decisions based off of it.
3) I think there are a lot of factors left out of the data set. For example, many times when someone leaves the company, it has very little to do with the company itself but rather with their personal life. Maybe the employee just had a child or their spouse got a job in another city. Because these aren't really represented in your data set, I think the model may end up not being totally accurate.