Where would you go this weekend?...

This is a created article in which the information may have evolved or changed. 1. [ABSTRACT] Time sensitivity is extremely important, the recommended time is too late or too late can be. , it is particularly important to predict the relatively distant activities of the future (which seems to be more difficult) than it is to predict the upcoming evil activities. This paper deals with the long-term processing of time-related methods. Based on a 4square dataset. The ability to achieve a given user's history predicts whether a user can access a particular location type on a given day. The forecast accuracy is 75% in a few weeks. Discover the fact: the interaction between points, the access history of this point can determine its future access points. 2.[introduction] predict that users will buy tickets in advance next week and buy them from here, many articles devoted to relevance, this article is devoted to timeliness. Forecasts are divided into short and long term and short term means "sudden" or "next activity". Short-term forecasts can be useful in applications such as transportation planning, but because users make many plans ahead of time, sometimes short-term plans are powerless. Long-term forecasts are necessary in such things as non-refundable air tickets and dinner set up by many people. 4.[technical approach] uses a classic two-dollar classification framework that produces a data point for every user, user visited location type, and active user history time. The label is "+1" or "1".      The following uses a series of features to calculate the description of each point, the data is divided into training sets and test sets, sorted by chronological order, the problem is that users sometimes change behavior patterns. The binary classifier is tested on the training set and evaluated on the test set. For each of these classifications, do the ROC curve on the err point used to evaluate the performance of the classification. The feature set used (not necessarily all in the forecast): Day of Week: the week in which the target is in the week (helps to capture the loop feature within a week) days since: Indicates the distance from the last visit to the destination history: is the feature used by each location type. Includes all access types, not just the target type. This feature reveals whether the site type was visited on the day related to the date that the user and the target dates are related. The selection of historical dates involves the following two points: A.today: The Access collection for the target date, capturing the current user activity, except for access to the target type B. N days ago:n access to all target location types