metrics是MAP@5,mean average precision
1. Leakage solution
2. most popular solution
Predict hotel type with pandas
因为每个srch_destination_id会对应N个hotel_cluster,所以根据bookings和clicks进行加权(weights=[1, 0.05]),得到relevance,然后得到训练集上srch_destination_id对应最大的5个hotel_cluster,然后将这个结果根据srch_destination_idmerge到测试集上。
只是热度推荐,做不到个性化推荐