"... In PAGE 2: ... Example 1. Consider a ten-record, five-dimensional customer dataset shown in Table1 . We are interested in dimensions Age-range, Car, and Salary-level, which may be useful for analyzing the latent behavior of the customers.... ..."

"... In PAGE 5: ... Example 1. Consider a ten-record customer dataset shown in Table1 . We are interested in dimensions Age-range, Car, and Salary-level, which are useful for analyzing the latent behavior of the customers.... In PAGE 5: ...5, we can get the set of all frequent patterns as shown in Table 2. According to formula (1), we can get the outlier factor value of each record as shown in Table1 (the fifth column). For each record, we also detect its top 1 contradict frequent patterns.... In PAGE 6: ...imilarly, the contradict-ness of {High} to this record is also 0.5. Hence, we list both of them. The top 1 contradict frequent patterns are listed in Table1 (the 6th column). According to the outlier definition, we can produce the top-5 outlier candidates with respect to the FPOF values.... ..."

"... In PAGE 11: ...uasi-extinction level of one is reasonable. Moreover, the Dennis et al. analyses, like most extinction models, neglect catastrophes and therefore tend to underestimate risks. For salmonids, catastrophes that could have a major impact on recruitment, such as major floods and debris flows, appear to occur as often as once every 100 years, but too infrequently to be represented in typical 10-15 year slices of data (see Table1 in Bisson et al .... ..."

"... In PAGE 5: ... Finally, tiers 4 and 5 include ASes of small regional ISPs and customer ASes respectively. As shown in Table2 , we have very good coverage of the top 4 AS tiers, with complete coverage of tier 1 and nearly-complete coverage of tier 2. 4 Confirming Anomalies Having collected the passive data from MonD and the traceroutes from ProbeD, the next step is processing the probes to confirm the existence of the anomaly.... ..."

"... In PAGE 5: ... Finally, tiers 4 and 5 include ASes of small regional ISPs and customer ASes respectively. As shown in Table2 , we have very good coverage of the top 4 AS tiers, with complete coverage of tier 1 and nearly-complete coverage of tier 2. 4 Confirming Anomalies Having collected the passive data from MonD and the traceroutes from ProbeD, the next step is processing the probes to confirm the existence of the anomaly.... ..."

"... In PAGE 8: ... In the rest of this section, we will explain our methodology with a toy example. Given the customer purchase data (such as in Table1 ), association mining, also known as market basket analysis, discovers the rules (relationships) that exist within the historical customer purchase (service order) data. Each of the rules include \if quot; clauses by default and the structure of the rules is as follows: If the customer purchases Product A, then with C% probability he/she will buy Product B.... In PAGE 8: ... The support level is the most important factor when pruning the association rule base. To depict our analysis with a simple example, a toy problem is provided in Table1 . USOCs given in Table 1 are actual values extracted from our databases.... In PAGE 8: ... To depict our analysis with a simple example, a toy problem is provided in Table 1. USOCs given in Table1 are actual values extracted from our databases. The descriptions of the USOCs are found in Table 2.... ..."

"... In PAGE 8: ... In the rest of this section, we will explain our methodology with a toy example. Given the customer purchase data (such as in Table1 ), association mining, also known as market basket analysis, discovers the rules (relationships) that exist within the historical customer purchase (service order) data. Each of the rules include if clauses by default and the structure of the rules is as follows: If the customer purchases Product A, then with C% probability he/she will buy Product B.... In PAGE 8: ... The support level is the most important factor when pruning the association rule base. To depict our analysis with a simple example, a toy problem is provided in Table1 . USOCs given in Table 1 are actual values extracted from our databases.... In PAGE 8: ... To depict our analysis with a simple example, a toy problem is provided in Table 1. USOCs given in Table1 are actual values extracted from our databases. The descriptions of the USOCs are found in Table 2.... ..."

"... In PAGE 6: ... Qualitative questions were scored using a rank system similar to the quantitative questions, on a scale of 1 to 3, for scoring of questions warranting a low, medium, or high response. Table5 illustrates the results of the customer-oriented variables for the respondents. For the discussion, qualitative data will be included to show reasoning of the sample organizations.... ..."

"... In PAGE 5: ...n application, cf. e.g. [ApRi2000]. To help with the distinction we propose the collection of characteristics in Table1 . Note that the given characteristics in both columns shall not be con- sidered as black-and-white, but rather as the opposite ends of a continuous spectrum.... ..."