A demo of forward stagewise regression.
This is a basic demo of forward stagewise regression. The model is used to predict the call time of two people, who are using HOLLA app to chat with each other.
The dataset are generated randomly by the gen_data.py
-
raw_data_diff_gender.csv: This dataset is used to predict the call time of a user when he or she is going to chat with a person of different gender.
-
raw_data_same_gender.csv: This dataset is used to predict the call time of a user when he or she is going to chat with a person of same gender.
The first 7 columns in each dataset are features selected to make prediction. These features are selected according to the a paper and the experience. The last column is the actual call time generated by human.
The selected features are: call_time_same_sex, call_time_diff_sex, quality_internet, time, if_new_friend, Credit, age_diff, language_fluence.
Description:
- call_time_same_sex: Average time of the call time when a user chats with person of the same sex, described by number of minutes.
- call_time_diff_sex: Average time of the call time when a user chats with person of the different sex, described by number of minutes.
- quality_internet : The quality of the Internet when a user begin chating, described by a score.
- time: This feature reflects if it is a comfortable time period for the user to chat, described by a score.
- if_new_friend: This feature reflects if a user begin chatting with a new friend, described by 0 or 1.
- Credit: This feature reflects the credit condition of a user, decribed by a score.
- age_diff: This feature reflects the age difference of the two users, decribed by number of years.
- language_fluence: This feature reflects the languge knowledge of a user, that is, if he or she can communicate well in the current language, decribed by scores.
Python 2.7 with numpy
Download and store the prediction_model.py and datasets in the same directory. And run the prediction_model.py. Then you will get the results.