# Problem Statement:
A large company having around 4000 employees. However, every year, around 15% of its employees leave the company and need to be replaced with the talent pool available in the job market. The management believes that this level of attrition (employees leaving, either on their own or because they got fired) is bad for the company, because of the following reasons -
1. The former employees’ projects get delayed, which makes it difficult to meet timelines, resulting in a reputation loss among consumers and partners
2. A sizeable department has to be maintained, for the purposes of recruiting new talent
3. More often than not, the new employees have to be trained for the job and/or given time to acclimatise themselves to the company
Hence, the management wants to understand what factors they should focus on, in order to curb attrition. In other words, they want to know what changes they should make to their workplace, in order to get most of their employees to stay. Also, they want to know which of these variables is most important and needs to be addressed right away.
# About Data:
Data is divide in to five files-
1. Employee information (Demographic information, education details, salary etc.)
2. Employee survey details
3. manager Survey Details
4. Employee in-time(check-in time)
5. Employee out-time(check-out-time)
# Solution:
As a part of the solution, I've done the
1. Collating all the files in to one
2. Data preparation
- Cleaning data
- checking for missing or NA values
- oulier detection/treatment
- Correcting the variable format
3. Exploratory Data Analysis (by plotting different graphs)
4. Feature standardisation
- Normalising continuous features
- Created the dummy variables for the categorical features
5. Divided the data in to the training and the test data
6. Build a Logistic regression model using the glm() function in R and also need to set the "family" parameter in function, it was binomial in our case.
- First build a general model with all the variables
- Then used the StepAIC function to remove the insignificant variables (direction as "both")
- I've checked for the p-value (which should be less than 0.05) and VIF (usually around 2 or 3 but we can consider variables with high value if p-value is significant or we can also try to remove the variable with high VIF and significant p-value, if it doesn't decrease the accuracy then we're good to remove those variables) for further removing the variables
7. Predicted the probabilities of Attrition for test data (as the output of the Logistic regression model is the probablities not the class)
8. Tested for the various probability cutoff (like 50%, 40% etc.) [in order to convert the proabalities in to class] and checked how many records are classified as correct by confusionMatrix.
9. Then find the optimal probalility cutoff and calculated the acuracy, specificity and sensitivity
accuracy - yeses (or positives) correctly predicted by it as yeses (or positives) and nos (or negatives) correctly predicted by the model as nos (or negatives)
- **Specificity -** specificity is equal to the proportion of nos (or negatives) correctly predicted by the model as nos (or negatives).
- **Sensitivity -** Sensitivity of a model is the proportion of yeses (or positives) correctly predicted by it as yeses (or positives).
10. Also, calculated the KS-statistis and plotted the Gain and Lift chart
* **KS-statistics:** A high KS statistic means that not only does your model have all churns at the top, it has has all non-churns at the bottom. For a good model, KS statistic would be more than 40% and would lie in the top few deciles
* **ROC curve:** I plotted ROC curve which can be plotted between % of bad and % of good, or in simple language, % of eventhappen and % of non-event happen.The perfect model is pretty much a right triangle, whereas the random model is a straight line. Basically, a model that rises steeply is a good model.
* **Gain and Lift Chart:** Gain and Lift chart are mainly concerned to check the rank ordering of the probabilities
# Which variables are indicators/predictor of attrition?
1. We’d seen employees in early 20’s and Single or having less experience are more prone to the attrition, So we have to emphasis on them either my meeting, survey’s etc.
2. Keep an eye on the employee past companies work duration. Make sure it’s not frequent.
3. Recently promoted
4. working environment and job satisfaction
Data-Science:R中的EDA和机器学习模型（回归，分类，聚类，SVM，决策树，随机森林，时间序列分析，推荐系统，XGBoost） （255个子文件） 
Amaze Payment Solution_EDA.pdf 1.16MB
APS_Analysis.md 27KB
unnamed-chunk-22-1.png 6KB unnamed-chunk-21-1.png 10KB unnamed-chunk-18-1.png 9KB unnamed-chunk-39-1.png 10KB unnamed-chunk-30-1.png 11KB unnamed-chunk-34-1.png 8KB unnamed-chunk-17-1.png 9KB unnamed-chunk-33-1.png 10KB unnamed-chunk-29-1.png 10KB unnamed-chunk-32-1.png 10KB unnamed-chunk-38-1.png 11KB unnamed-chunk-35-1.png 8KB unnamed-chunk-42-1.png 10KB unnamed-chunk-41-1.png 8KB unnamed-chunk-40-1.png 10KB unnamed-chunk-31-1.png 10KB unnamed-chunk-24-1.png 8KB unnamed-chunk-36-1.png 11KB unnamed-chunk-26-1.png 9KB unnamed-chunk-27-1.png 10KB unnamed-chunk-28-1.png 10KB unnamed-chunk-25-1.png 9KB unnamed-chunk-19-1.png 9KB unnamed-chunk-23-1.png 9KB unnamed-chunk-15-1.png 8KB unnamed-chunk-37-1.png 10KB unnamed-chunk-20-1.png 10KB README.md 3KB APS_Analysis.R 21KB unnamed-chunk-12-1.png 8KB unnamed-chunk-33-1.png 22KB unnamed-chunk-29-1.png 9KB unnamed-chunk-38-1.png 11KB unnamed-chunk-16-1.png 9KB unnamed-chunk-35-1.png 12KB unnamed-chunk-40-1.png 10KB unnamed-chunk-14-1.png 24KB unnamed-chunk-26-1.png 10KB unnamed-chunk-58-2.png 31KB unnamed-chunk-37-1.png 12KB unnamed-chunk-58-1.png 31KB Beer_Recommendation.R 17KB Beer_Recommender_System.md 30KB README.md 2KB README.md 8B Telecom_churn_Model.md 102KB Telecom_churn_Model.R 16KB unnamed-chunk-8-1.png 18KB unnamed-chunk-7-1.png 41KB unnamed-chunk-12-1.png 12KB unnamed-chunk-13-1.png 12KB unnamed-chunk-9-1.png 30KB unnamed-chunk-11-1.png 10KB unnamed-chunk-44-1.png 31KB unnamed-chunk-14-1.png 12KB unnamed-chunk-15-1.png 23KB churn_data.csv 480KB README.md 42B customer_data.csv 184KB Telecom Churn Data Dictionary.csv 1KB internet_data.csv 456KB README.md 2KB Ads.R 10KB README.md 2KB README.md 29B Letter Recognition.R 3KB README.md 4KB mnist_test.csv 17.44MB README.md 17B Loan_Defaulter.R 20KB Loan_Defaulter.md 21KB Loan_Defaulter.pdf 1.75MB unnamed-chunk-51-1.png 97KB unnamed-chunk-22-1.png 27KB unnamed-chunk-18-1.png 57KB unnamed-chunk-30-1.png 27KB unnamed-chunk-46-1.png 22KB unnamed-chunk-34-1.png 30KB unnamed-chunk-32-1.png 28KB unnamed-chunk-38-1.png 51KB unnamed-chunk-16-1.png 26KB unnamed-chunk-42-1.png 37KB unnamed-chunk-44-1.png 11KB unnamed-chunk-53-1.png 27KB unnamed-chunk-48-1.png 11KB unnamed-chunk-40-1.png 105KB unnamed-chunk-24-1.png 30KB unnamed-chunk-59-1.png 68KB unnamed-chunk-36-1.png 27KB unnamed-chunk-57-1.png 20KB unnamed-chunk-26-1.png 42KB unnamed-chunk-55-1.png 22KB unnamed-chunk-28-1.png 36KB unnamed-chunk-20-1.png 30KB README.md 4KB README.md 42B
Data_Dictionary.xlsx 18KB Credit_Card_Defaulter.R 1KB共 255 条
- 1
- 2
- 3
评论0