Electrical computers and digital processing systems: multicomput – Distributed data processing – Client/server
Reexamination Certificate
1998-04-24
2001-04-10
Black, Thomas G. (Department: 2771)
Electrical computers and digital processing systems: multicomput
Distributed data processing
Client/server
C709S217000, C709S223000, C707S793000
Reexamination Certificate
active
06216154
ABSTRACT:
§1. BACKGROUND OF THE INVENTION
§1.1 Field of the Invention
The present invention concerns time dependency of certain variables such as, for example, the time dependency of the use of (e.g., requests for) Internet resources. In particular, the present invention concerns entering and evaluating time dependency hypotheses and forecasting variable values based on the time dependency hypotheses.
§1.2 Related Art
People rely on forecasting, of one type or the other, in many ways. For example, people may rely on weather forecasts to determine what to wear or what crops to plant, people may rely on forecasts of a stock market index or of a particular company's earnings when investing money, and companies may rely on demand forecasts when deciding what and how many products to produce. Forecasting or predictions also extend to less essential issues, such as what team will win this week's football game or what movie will win this year's Academy Award.
Some forecasts are made by detecting temporal patterns in known data and extrapolating the data forward based on the detected patterns. For example, a forecast of the number of people making a telephone call during a particular time period in a day may be made based on detected temporal patterns in historical data of the number of people making telephone calls during that particular time period in past days.
One general purpose of forecasting is to predict what people will want in the future so that those wants can be met. For example, in the context of networked computers, such as the Internet, resource servers (also referred to as “web sites” or “Internet sites”) service requests for content (e.g., documents, HTML (“Hyper Text Mark-up Language”) pages, JPEG (“Joint Photographic Experts Group” encoded) images, MPEG (“Motion Picture Experts Group” encoded) video, audio information, etc.) from clients. If client requests can be accurately forecast, resource servers may be adapted to service such client requests in a more efficient (e.g., in terms of ease of navigation, download time, availability, etc.) manner.
Forecasts of requests for Internet resources may be made based on the resource itself, or one or more attributes of the resources. Different Internet resources may have different attributes. For example, an Internet resource providing a stock market report may have “FINANCIAL” and “HTML” attributes, an Internet resource providing an image of a famous painting may have “ART” and “JPEG” attributes, and a video clip of baseball highlights may have “SPORTS” and “MPEG” attributes.
FIG. 1
a
is a notional plot
110
of requests for an Internet resource which provides an interactive game, over time. As shown in
FIG. 1
a
, requests may increase after working hours and increase dramatically (See
115
a
and
115
b
) during the weekends.
FIG. 1
b
is a notional plot
120
of requests for an Internet resource which provides a download for a computer program, over time. If the download is provided for free during a certain time period, a spike
125
in requests may occur during that time period.
FIG. 1
c
is a notional plot
130
of requests for an Internet resource having content which is updated every Wednesday. As shown in
FIG. 1
c
, requests may increase on Wednesdays (See
135
a
and
135
b
) and taper off for the rest of the week.
As can be seen from the above examples, the temporal properties of Internet resource requests may depend on the attributes of those resources. If the number of attributes is relatively small, such patterns might be readily discernable merely by looking at temporal patterns in the plots of data (e.g., requests). However, large Internet sites may have resources with thousands of attributes. Thus, a tool is needed to automatically forecast requests for Internet resources.
One of the most popular time series models used in forecasting is the seasonal ARIMA (or Auto-Regressive Integrated Moving Average) model. Unfortunately, however, a fundamental assumption of the ARIMA model makes it unsuitable for forecasting events which exhibit non-homogeneous time intervals (e.g., different patterns on weekdays and weekends). It is believed that requests for Internet resources are not time-homogeneous. Thus, a tool for forecasting events that are not time-homogeneous is needed.
§2. SUMMARY OF THE INVENTION
The present invention provides a user interface for permitting non-time-homogeneous forecast hypotheses to be entered. The user interface permits the entry of groups of similar dependence hypotheses. A dependence hypothesis may be a pair of variables (e.g., times) where a value associated with the second variable (e.g., time) is expected to be significantly dependent on a value associated with the first variable (e.g., time). Similar dependence hypotheses are expected to have similar dependence over all of the hypotheses. In the context of the Internet for example, the number of requests for a resource on Tuesday night may be expected to depend on the number of requests for that resource on Monday night, and the number of requests for the resource on Wednesday night may be expected to similarly depend on the number of requests for the resource on Tuesday night.
The present invention also tests the groups of similar dependence hypotheses against historical data and to provide a indication of the validity of the groups of similar dependence hypotheses. More specifically, the present invention may present a display of (i) a scatter diagram of historical data related to a group of similar dependence hypotheses, (ii) correlations of various groups of similar dependence hypotheses to historical data, and (iii) forecasts of values.
Further, the present invention also determines how various groups of historical data (e.g., requests for various resources or resources having various attributes) fit a given group(s) of similar dependence hypotheses.
Finally, the present invention provides a tool which permits a user to interactively explore data to discover unforeseen patterns.
REFERENCES:
patent: 5452440 (1995-09-01), Salsburg
patent: 5727129 (1998-03-01), Barrett et al.
patent: 5802292 (1998-09-01), Mogul
patent: 5924116 (1999-07-01), Aggarwal et al.
Markatos (“A Top-10 Approach to Prefetching on the Web”, Institute of Computer Science Aug. 1996, pp. 1-15).*
Grimsrud (“Multiple prefetch adaptive disk caching”, IEEE Transactions on Knowledge and Data Engineering, vol. 5, No. 1, Feb. 1993, pp. 88-103)
Altschuler Steven J.
Ingerman David
Black Thomas G.
Michaelson Peter L.
Michaelson & Wallace
Microsoft Corporation
Trinh William
LandOfFree
Methods and apparatus for entering and evaluating time... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Methods and apparatus for entering and evaluating time..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Methods and apparatus for entering and evaluating time... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2487294