Embedded Files
Short description of our dataset
Short description of our dataset
The topic of the second dataset is about churn in Telecom. We found on Kaggle, a website where we can work on data science project.
The topic of the second dataset is about churn in Telecom. We found on Kaggle, a website where we can work on data science project.
The main purpose is to predict if a customer will soon churn his subscribe.
The main purpose is to predict if a customer will soon churn his subscribe.
Why this dataset ?
Why this dataset ?
This dataset might be suitable because first of all it has been given by the French government, so it seems legit to assume that it should be a good quality dataset.
This dataset might be suitable because first of all it has been given by the French government, so it seems legit to assume that it should be a good quality dataset.
It could also be good in order to take care of our different data mining benefits, because we can find the quality of the network (coverage) depending on many characteristics (operation, location …)
It could also be good in order to take care of our different data mining benefits, because we can find the quality of the network (coverage) depending on many characteristics (operation, location …)
The coverage will be our most important data, mainly because it gave us the quality of the data which is the main point of our different analysis.
The coverage will be our most important data, mainly because it gave us the quality of the data which is the main point of our different analysis.
Then, we can also remark that there is some missing data. These latter are paradoxically essential for the cover and when the network is 4G. It is because the 4G network was still not available all over the France by July 2015, so it is normal that there is no cover at all for some city.
Then, we can also remark that there is some missing data. These latter are paradoxically essential for the cover and when the network is 4G. It is because the 4G network was still not available all over the France by July 2015, so it is normal that there is no cover at all for some city.
Those missing data are normal so there is nothing we can do about it, except including them in our different analysis.
Those missing data are normal so there is nothing we can do about it, except including them in our different analysis.
We can find the different information:
We can find the different information:
- The state. This field is nominal.
- The account length. This field is nominal.
- The area code. This field is nominal.
- The phone number. This field is nominal.
- The international plan. This field is binary.
- The voice mail plan. This field is binary.
- The number vmail messages. This field is quantity.
- The total day minutes. This field is quantity.
- The total day calls. This field is quantity.
- The total day charges. This field is quantity.
- The total eve minutes. This field is quantity.
- The total eve calls. This field is quantity.
- The total eve charge. This field is quantity.
- The total night minutes. This field is quantity.
- The total night calls. This field is quantity.
- The total night charge. This field is quantity.
- The total intl minutes. This field is quantity.
- The total intl calls. This field is quantity.
- The total intl charges. This field is quantity.
- The customer service calls. This field is quantity.
- The churn. This field is binary.
Screenshot example of the dataset :
Screenshot example of the dataset :
For your information
For your information
At the beginning our group had another dataset. However after some analysiswe were not satisfied of the results because there was no corelation between the diferents elements of our dataset, there was no real problem to draw from this dataset and the results were totally ilogical even when we apllied the advices of the teacher. So we tried with this new dataset in order to be sure that we did the things well and finally the results were much better than the first dataset.
At the beginning our group had another dataset. However after some analysiswe were not satisfied of the results because there was no corelation between the diferents elements of our dataset, there was no real problem to draw from this dataset and the results were totally ilogical even when we apllied the advices of the teacher. So we tried with this new dataset in order to be sure that we did the things well and finally the results were much better than the first dataset.
Google Sites
Report abuse