The hotel chain IMSInn, like many other hotels, has been keeping score of all the customer satisfaction surveys their guest kindly fill out and leave in their room after their stay. However, until know they were only used to obtain an overall satisfaction score.
\nRecent developments of the global pandemic have caused business to slow down and the marketing department is working like crazy to figure a strategy to keep the impacts at a minimum. The idea is to target customers with a high probability of returning and give them discounts if they make a booking for next year, helping to contain marketing costs while guarantying some cash inflow.
\nYour goal is to build a predictive model that answers the question “Which guests are more likely to return?” using the small quantity of data accessible from the guest data base that contains general information about the guests and their satisfaction survey responses, explained bellow.
\n\n<p>\nThe training set should be used to build your machine learning models. In this set, you also have the ground truth associated with each guest, i.e., if he returned or not.\nThe test set should be used to see how well your model performs on unseen data. In this set you don’t have access to the ground truth, and the goal of your team is to predict that value (“0” or “1”) by using the model you created using the training set.\nThe score of your predictions is calculated using f1 score.\n</p>\n\n
<p>\nYou should submit a csv file with the number of instances in the test set, containing the columns [Guest_ID, Repeater], and only those columns. The Repeater column should contain the prediction (“0” or “1”).\n</p>", "isPublished": false, "hasIsPublished": false, "postTitle": "", "hasPostTitle": false, "mimeType": "", "hasMimeType": false }, { "name": "Evaluation", "content": "
The evaluation metric for this competition is Mean F1-Score. The F1 score, commonly used in information retrieval, measures accuracy using the statistics precision p and recall r. Precision is the ratio of true positives (tp) to all predicted positives (tp + fp). Recall is the ratio of true positives to all actual positives (tp + fn). The F1 score is given by:
\n\n\\[ F1 = 2\\frac{p \\cdot r}{p+r}\\ \\ \\mathrm{where}\\ \\ p = \\frac{tp}{tp+fp},\\ \\ r = \\frac{tp}{tp+fn} \\]
\nThe F1 metric weights recall and precision equally, and a good retrieval algorithm will maximize both precision and recall simultaneously. Thus, moderately good performance on both will be favored over extremely good performance on one and poor performance on the other.
\nFor every guest in the dataset, submission files should contain two columns: Guest_ID and Repeater. \n
The file should contain a header and have the following format:
\nGuest_ID,Repeater", "isPublished": false, "hasIsPublished": false, "postTitle": "", "hasPostTitle": false, "mimeType": "", "hasMimeType": false }, { "name": "rules", "content": "**NOTE:** The deliverables / submissions / projects that do not have the below conditions will be highly penalized.\n\n**ABOUT THE PROJECT**
19847,1
12433,0
10273,1
etc.