"1. On average, each student has 80.52 (SD=170.17) files and 13.58 (SD=45.33) directories. Table 1. Descriptive statistics for the four describing variables. After clustering the students according to the four variables, we have calculated means and SD for each variable within each cluster; results are given in Table 2, where maximum and minimum values for each variable are bolded and italicized, accordingly. Table 2. Means (SD) of the four variables by which the clusters were formed. As might be seen from the table, Cluster 1 (n=141) is characterized by extreme values of two variables' means among clusters: Pile rate gets a maximum (0.97), and inner-pile rate gets a minimum (0.02). These results imply that in this cluster, most of the students' files are stored in the root directory (hence it is not surprising that the second largest folder is extremely small). These two extreme values of variables are typical for Piling organization strategy. In Cluster 2 (n=49), again the means of the same two variables as in Cluster 1 get to their extreme values, however in different direction. In this cluster, the mean of pile rate is minimal (0.09), and we may think that this is a non-piling strategy. However, the mean of inner-pile rate is relatively high (0.86), which indicates on the existence of a folder holding a large share of the archive. That means that the files were saved in one main folder out of the root directory – a strategy that we may call One-folder Filing. Cluster 3 (n=262) has minimum mean values for two variables: Files per folder and Largest folder, i.e., students in it have small folders on average (6.1), and their largest folder is also relatively small (14.52). This suggests that the cluster represents a Small- folders Filing organization strategy. In Cluster 4 (n=66), the means of the same two variables as in Cluster 3 take their extreme values: Both files per folder (23.1) and largest folder (71.62) are maximal. By examining the mean value of pile rate (0.13), it might be concluded that about 87% of their files are filed, with one folder containing about half of their files (0.48). Therefore, this cluster, which we call Big-folder Filing, describes a mixture of filing and piling. According to this analysis of the clusters, we present the following classification of personal information space organization strategies: Piling, One-folder Filing, Small- folders Filing, and Big-folder Filing. Table 3 shows the distribution of the four types in the research population. Table 3. Personal Information Space Organization Strategies distribution. For examining the association between the archive size and its organization strategy, mean values for archive size (total number of files) were compared between the clusters. Using Univariate ANOVA test, it was shown that the means are significantly different. As may be seen from Table 4, two strategies (Piling, One-folder Filing) have a small archive size on average (24.4 and 22.31, respectively), while the largest mean value for archive size (284.73) was found in the Big-folder Filing cluster. This indicates that larger archives are associated with strategies of filing into more than one directory. Table 4. Archive size in the different clusters. 5 Discussion. The main purpose of this study was to empirically identify different types of personal information organization strategies, which are part of Personal Information Management (PIM), and to do so for a large population, using data mining methodologies. PIM is not only a coherent and integral part of the learning process in the digital era - it is a process through which students learn. Therefore, researching PIM in the context of learning is very important for having a broader understanding of the learning process. Applying data mining techniques for PIM research brings new and fascinating opportunities to this field, as was demonstrated in this study. Focusing on users' management of online personal archives, we were able to empirically identify four types of archiving strategies: a) Piling – most of the files are in the root directory; b) One-folder Filing – most of the files are located in one folder, under the root directory; c) Small Folders Filing – items are being divided into many relatively small folders (about 6 files per folder on average); d) Big-folder Filing – items are being divided into folders (about 23 files per folder on average) with about a half of them located in one big folder. These four types refine the classical Filing/Piling binary classification [14]. As the results suggest, students who tend to be Big-folder Filers, manage the largest archives and have relatively many files per folder on average. In order to construct a hierarchy of large coherent folders of different items related to a certain context (represented by each folder's name), students are required to a meaningful integration and generalization processes regarding the subject matter. Our analysis showed that more than half of the participating students were categorized as Small-folders Filers. As this strategy is characterized by the use of small folders, this might imply that there are relatively many near-empty folders. Empty folders might indicate on a pre-building strategy, as was previously observed in the context of students' PIM [8]. Having many empty folders might increase PIM complexity, as well as having big folders. The strategy of Big-folder Filing was found in this study as associated with large archives, supporting previous findings [11]. In the context of learning, increasing PIM complexity is of special interest as PIM activities require cognitive skills. Bloom's cognitive taxonomy for learning objectives [5] enables us to analyze the three main PIM activities – i.e., naming, sorting, and categorization – in the light of three different levels of the taxonomy's cognitive skills: knowledge, analysis, and synthesis, accordingly. Regarding the four personal organization strategies found in this study, we might suggest different levels of reflected activities. In Piling strategy, the students neither name, sort nor categorize any information items. In One Folder Filing strategy, the students name only few folders and don't sort or categorize at all. In Small Folders Filing, the students name folder and sort information items into them, however they only do little categorization (since they join only few items into each folder). Only in Big-folder Filing strategy, students name, sort and categorize many items into folders. As the results suggest, managing bigger archives requires a wider range of cognitive skills. Replicating the process described in this article over several points in time might enlighten issues regarding changes over time of PIM strategies and their related cognitive activities. PIM is subjective and idiosyncratic, and because PIM research mostly uses qualitative data collection from relatively small populations, it might seem that there are as many PIM variations as there are researched users [12]. However, using a large research population and data mining techniques, unexpected patterns might arise, suggesting similarities between groups of users, as was shown in this study. To promote the creation of large datasets, Chernov et al. [9] have suggested building a repository of PIM activity log files; this then would serve the PIM research community. Since it is likely that there will be problems obtaining participants' consent to trace their PIM activity over time, it might be easier to collect structural data reflecting accumulating activity."