Method and system for predicting user activity levels associated with an application
First Claim
1. A method of predicting a user activity level associated with an application executing on a computing system in a multi-user computing environment, comprising:
- defining a user activity metric (UAM) as a variable, said UAM capable of being assigned a plurality of activity levels, wherein each activity level of said plurality of activity levels is a measured number of transactions performed by a plurality of users utilizing said application during a time period of a plurality of time periods, or a measured number of said users utilizing said application during said time period;
defining a plurality of factors associated with said plurality of activity levels, said plurality of factors including a calendar type factor, said calendar type factor capable of being assigned a plurality of values of a cyclic time scale, wherein each value of said plurality of values represents an amount of time elapsed from a specified event;
assigning said plurality of activity levels to said UAM;
assigning said plurality of values to said calendar type factor of said plurality of factors, said plurality of values associated with said plurality of activity levels in a one-to-one correspondence, and with said plurality of time periods in a one-to-one correspondence;
adjusting, subsequent to said assigning said plurality of values to said calendar type factor, said plurality of values to a plurality of adjusted values, said plurality of adjusted values including a pre-defined base value and one or more other adjusted values, wherein an adjusted value of said one or more other adjusted values represents a distance to said pre-defined base value, and wherein said adjusting includes facilitating a representation of a linear dependency between said plurality of adjusted values and said plurality of activity levels;
calculating a plurality of coefficients of correlation, said calculating said plurality of coefficients of correlation including calculating a first coefficient of correlation between said plurality of adjusted values and said plurality of activity levels;
determining that an absolute value of said first coefficient of correlation is less than a pre-defined threshold value;
excluding, in response to said determining, said calendar type factor from said plurality of factors, said excluding including forming a subset of said plurality of factors, said subset including one or more factors, wherein an absolute value of a second coefficient of correlation between a set of values assigned to any factor of said one or more factors and said plurality of activity levels is greater than or equal to said pre-defined threshold value; and
generating a regression model to predict an activity level, said regression model based on said plurality of activity levels and said subset.
1 Assignment
0 Petitions
Accused Products
Abstract
A method and system for predicting a user activity level associated with an application. An activity level is a number of transactions performed by users utilizing the application per time period or a number of users utilizing the application per time period. Measurements of activity levels are assigned to a user activity metric (UAM) variable, and associated values are assigned to a set of factors. At least one correlation coefficient between each factor and the UAM is calculated. In response to a maximum correlation coefficient associated with a factor being less than a pre-defined threshold, the factor is excluded from the set of factors to facilitate forming a subset of factors associated with correlation coefficients whose absolute values are greater than or equal to the pre-defined threshold. A regression model utilizing the subset is generated to predict an activity level.
-
Citations
40 Claims
-
1. A method of predicting a user activity level associated with an application executing on a computing system in a multi-user computing environment, comprising:
-
defining a user activity metric (UAM) as a variable, said UAM capable of being assigned a plurality of activity levels, wherein each activity level of said plurality of activity levels is a measured number of transactions performed by a plurality of users utilizing said application during a time period of a plurality of time periods, or a measured number of said users utilizing said application during said time period;
defining a plurality of factors associated with said plurality of activity levels, said plurality of factors including a calendar type factor, said calendar type factor capable of being assigned a plurality of values of a cyclic time scale, wherein each value of said plurality of values represents an amount of time elapsed from a specified event;
assigning said plurality of activity levels to said UAM;
assigning said plurality of values to said calendar type factor of said plurality of factors, said plurality of values associated with said plurality of activity levels in a one-to-one correspondence, and with said plurality of time periods in a one-to-one correspondence;
adjusting, subsequent to said assigning said plurality of values to said calendar type factor, said plurality of values to a plurality of adjusted values, said plurality of adjusted values including a pre-defined base value and one or more other adjusted values, wherein an adjusted value of said one or more other adjusted values represents a distance to said pre-defined base value, and wherein said adjusting includes facilitating a representation of a linear dependency between said plurality of adjusted values and said plurality of activity levels;
calculating a plurality of coefficients of correlation, said calculating said plurality of coefficients of correlation including calculating a first coefficient of correlation between said plurality of adjusted values and said plurality of activity levels;
determining that an absolute value of said first coefficient of correlation is less than a pre-defined threshold value;
excluding, in response to said determining, said calendar type factor from said plurality of factors, said excluding including forming a subset of said plurality of factors, said subset including one or more factors, wherein an absolute value of a second coefficient of correlation between a set of values assigned to any factor of said one or more factors and said plurality of activity levels is greater than or equal to said pre-defined threshold value; and
generating a regression model to predict an activity level, said regression model based on said plurality of activity levels and said subset. - View Dependent Claims (2, 3, 6)
-
-
4. (canceled)
-
5. A method of predicting a user activity level associated with an application executing on a computing system in a multi-user computing environment, comprising:
-
defining a user activity metric (UAM) as a variable, said UAM capable of being assigned a plurality of activity levels, wherein each activity level of said plurality of activity levels is a measured number of transactions performed by a plurality of users utilizing said application during a time period of a plurality of time periods, or a measured number of said users utilizing said application during said time period;
defining a plurality of factors associated with said UAM;
assigning said plurality of activity levels to said UAM;
assigning a plurality of values to a factor of said plurality of factors, said plurality of values associated with said plurality of activity levels in a one-to-one correspondence and with said plurality of time periods in a one-to-one correspondence;
calculating at least one coefficient of correlation between at least two values of said plurality of values and at least two activity levels of said plurality of activity levels;
determining a maximum coefficient of correlation of said at least one coefficient of correlation that is less than a pre-defined threshold value, excluding, in response to said determining, said factor from said plurality of factors to facilitate forming a subset of said plurality of factors, said subset including one or more factors, wherein an absolute value of a coefficient of correlation between any factor of said one or more factors and said UAM is greater than or equal to said pre-defined threshold value;
generating a regression model to predict an activity level, said regression model based on said plurality of activity levels and said subset, wherein said defining said plurality of factors comprises defining said factor as a calendar type factor for which each value of said plurality of values represents an amount of time elapsed from a specified event, wherein said at least one coefficient of correlation comprises a single coefficient of correlation, and wherein said maximum coefficient of correlation is an absolute value of said single coefficient of correlation;
determining a sample mean, a standard deviation, and a standard error associated with said plurality of activity levels, said plurality of activity levels being historical data;
modifying a value of said plurality of values to a pre-defined base value, wherein said value is associated with an activity level of said plurality of activity levels exceeding said sample mean by more than said standard error multiplied by a pre-defined number; and
adjusting other values of said plurality of values to a set of two or more adjusted values, wherein an adjusted value of said two or more adjusted values represents a distance to said base value, wherein said calculating utilizes said base value and said set of two or more adjusted values to calculate said single coefficient of correlation, wherein a first absolute value of said single coefficient of correlation is greater than a second absolute value of another coefficient of correlation between said plurality of values and said plurality of activity levels. - View Dependent Claims (37, 38)
-
-
7. A method of predicting a user activity level associated with an application executing on a computing system in a multi-user computing environment, said method comprising:
-
defining a user activity metric (UAM) as a variable, said UAM capable of being assigned a plurality of activity levels, wherein each activity level of said plurality of activity levels is a measured number of transactions performed by a plurality of users utilizing said application during a time period of a plurality of time periods, or a measured number of said plurality of users utilizing said application during said time period;
defining a plurality of factors associated with said plurality of activity levels, said plurality of factors including a first subset of said plurality of factors, said first subset including at least one factor, wherein each factor of said first subset is a measurable variable on which said UAM depends, wherein said measurable variable is not a calendar type factor capable of being assigned amounts of time elapsed from a specified event;
assigning said plurality of activity levels to said UAM, said assigning said plurality of activity levels including assigning a plurality of sets of at least two activity levels, wherein each set of said at least two activity levels depends upon a factor of said plurality of factors;
assigning a plurality of values to each factor of said plurality of factors, said plurality of factors associated with said plurality of activity levels in a one-to-one correspondence, and with said plurality of time periods in a one-to-one correspondence;
generating a first matrix of xij elements, wherein xij is a value of an i-th factor of said first subset of said plurality of factors, said value of said i-th factor measured at an end of a j-th time period of said plurality of time periods;
generating a second matrix including said first matrix, wherein said second matrix is a three dimensional matrix including yijk elements, wherein yijk is an (i,j,k)-th activity level assigned to said UAM, and included in a k-th set of said plurality of sets of at least two activity levels, said (i,j,k)-th activity level associated with said xij and measured at an end of a (j+k)-th time period of said plurality of time periods;
calculating a plurality of coefficients of correlation, wherein each coefficient of correlation of said plurality of coefficients of correlation is rik and wherein said rik is an (i,k)-th coefficient of correlation between said i-th factor of said first subset of said plurality of factors and said k-th set of said plurality of sets of at least two activity levels;
determining that each absolute value of rik coefficients of correlation associated with said i-th factor is less than a pre-defined threshold value;
excluding, in response to said determining, said i-th factor from said plurality of factors, said excluding including forming a subset of said plurality of factors, said subset of said plurality of factors including one or more factors, wherein an absolute value of a coefficient of correlation between a set of values assigned to any factor of said one or more factors and said plurality of activity levels is greater than or equal to said pre-defined threshold value; and
generating a regression model to predict an activity level, said regression model based on said plurality of activity levels and said subset of said plurality of factors.
-
-
8. A method of predicting a user activity level associated with an application executing on a computing system in a multi-user computing environment, comprising:
-
defining a user activity metric (UAM) as a variable, said UAM capable of being assigned a plurality of activity levels, wherein each activity level of said plurality of activity levels is a measured number of transactions performed by a plurality of users utilizing said application during a time period of a plurality of time periods, or a measured number of said users utilizing said application during said time period;
defining a plurality of factors associated with said UAM;
assigning a plurality of activity levels to said UAM;
assigning a plurality of values to a factor of said plurality of factors, said plurality of values associated with said plurality of activity levels in a one-to-one correspondence, and with said plurality of time periods in a one-to-one correspondence;
calculating at least one coefficient of correlation between at least two values of said plurality of values and at least two activity levels of said plurality of activity levels;
determining a maximum coefficient of correlation of said at least one coefficient of correlation that is less than a pre-defined threshold value;
excluding, in response to said determining, said factor from said plurality of factors to facilitate forming a subset of said plurality of factors, said subset including one or more factors, wherein an absolute value of a coefficient of correlation between any factor of said one or more factors and said UAM is greater than or equal to said pre-defined threshold value; and
generating a regression model to predict an activity level, said regression model based on said plurality of activity levels and said subset, wherein said defining said plurality of factors comprises;
defining a first subset of said plurality of factors, said first subset including at least one factor, wherein said factor is included in said first subset, wherein each factor of said first subset is a measurable variable on which said UAM depends, wherein said measurable variable is not a calendar type factor for which each value of said plurality of values represents an amount of time elapsed from a specified event, and wherein said method further comprises;
assigning said plurality of activity levels to said UAM, said plurality of activity levels included in a plurality of sets of at least two activity levels, wherein each set of said at least two activity levels depends upon a factor of said plurality of factors, and is included in said plurality of activity levels, and wherein said plurality of activity levels is included in one set of said plurality of sets;
generating a first matrix of xij elements, wherein xij is a value of an i-th factor of said first subset measured at an end of a j-th time period of said plurality of time periods; and
generating a second matrix including said first matrix, wherein said second matrix is a three dimensional matrix including yijk elements, wherein yijk is an (i,j,k)-th activity level assigned to said UAM, and included in a k-th set of said plurality of sets of at least two activity levels, said (i,j,k)-th activity level associated with said xij and measured at an end of a (j+k)-th time period of said plurality of time periods, and wherein said second matrix includes elements rik, wherein rik is an (i,k)-th coefficient of correlation between said i-th factor of said first subset and said k-th set, wherein said maximum coefficient of correlation is an absolute value of an element (r) included in a set of said elements rik, wherein said absolute value of said r is greater than or equal to an absolute value of any other element of said set of said elements rik, said r being associated with said factor and with a set of said plurality of sets of at least two activity levels, and wherein said maximum coefficient of correlation is associated with a time step K, wherein each activity level of said set of said plurality of sets is measured K units of time after an end of a period of time of said plurality of periods of time. - View Dependent Claims (9, 39, 40)
-
-
10. A system for predicting a user activity level associated with an application executing on a computing system in a multi-user computing environment, comprising:
-
means for defining a user activity metric (UAM) as a variable, said UAM capable of being assigned a plurality of activity levels, wherein each activity level of said plurality of activity levels is a measured number of transactions performed by a plurality of users utilizing said application during a time period of a plurality of time periods, or a measured number of said users utilizing said application during said time period;
means for defining a plurality of factors associated with said plurality of activity levels, said plurality of factors including a calendar type factor, said calendar type factor capable of being assigned a plurality of values of a cyclic time scale, wherein each value of said plurality of values represents an amount of time elapsed from a specified event;
means for assigning said plurality of activity levels to said UAM;
means for assigning said plurality of values to said calendar type factor of said plurality of factors, said plurality of values associated with said plurality of activity levels in a one-to-one correspondence, and with said plurality of time periods in a one-to-one correspondence;
means for adjusting, subsequent to said assigning said plurality of values to said calendar type factor, said plurality of values to a plurality of adjusted values, said plurality of adjusted values including a pre-defined base value and one or more other adjusted values, wherein an adjusted value of said one or more other adjusted values represents a distance to said pre-defined base value, and wherein said adjusting includes facilitating a representation of a linear dependency between said plurality of adjusted values and said plurality of activity levels;
means for calculating a plurality of coefficients of correlation, said calculating said plurality of coefficients of correlation including calculating a first coefficient of correlation between said plurality of adjusted values and said plurality of activity levels;
means for determining that an absolute value of said first coefficient of correlation is less than a pre-defined threshold value;
means for excluding, in response to said determining, said calendar type factor from said plurality of factors, said excluding including forming a subset of said plurality of factors, said subset including one or more factors, wherein an absolute value of a second coefficient of correlation between a set of values assigned to any factor of said one or more factors and said plurality of activity levels is greater than or equal to said pre-defined threshold value; and
means for generating a regression model to predict an activity level, said regression model based on said plurality of activity levels and said subset. - View Dependent Claims (11, 12, 15)
-
-
13. (canceled)
-
14. (canceled)
-
16. A system of predicting a user activity level associated with an application executing on a computing system in a multi-user computing environment, said system comprising:
-
means for defining a user activity metric (UAM) as a variable, said UAM capable of being assigned a plurality of activity levels, wherein each activity level of said plurality of activity levels is a measured number of transactions performed by a plurality of users utilizing said application during a time period of a plurality of time periods, or a measured number of said plurality of users utilizing said application during said time period;
means for defining a plurality of factors associated with said plurality of activity levels, said plurality of factors including a first subset of said plurality of factors, said first subset including at least one factor, wherein each factor of said first subset is a measurable variable on which said UAM depends, wherein said measurable variable is not a calendar type factor capable of being assigned amounts of time elapsed from a specified event;
means for assigning said plurality of activity levels to said UAM, said means for assigning said plurality of activity levels including means for assigning a plurality of sets of at least two activity levels, wherein each set of said at least two activity levels depends upon a factor of said plurality of factor;
means for assigning a plurality of values to each factor of said plurality of factors, said plurality of factors associated with said plurality of activity levels in a one-to-one correspondence, and with said plurality of time periods in a one-to-one correspondence;
means for generating a first matrix of xij elements, wherein xij is a value of an i-th factor of said first subset of said plurality of factors, said value of said i-th factor measured at an end of a j-th time period of said plurality of time periods;
means for generating a second matrix including said first matrix, wherein said second matrix is a three dimensional matrix including yik elements, wherein yik is an (i,j,k)-th activity level assigned to said UAM, and included in a k-th set of said plurality of sets of at least two activity levels, said (i,j,k)-th activity level associated with said xij and measured at an end of a (j+k)-th time period of said plurality of time periods;
means for calculating a plurality of coefficients of correlation, wherein each coefficient of correlation of said plurality of coefficients of correlation is rik and wherein said rik is an (i,k)-th coefficient of correlation between said i-th factor of said first subset of said plurality of factors and said k-th set of said plurality of sets of at least two activity levels;
means for determining that each absolute value of rik coefficients of correlation associated with said i-th factor is less than a pre-defined threshold value;
means for excluding, in response to said determining, said i-th factor from said plurality of factors, said means for excluding including means for forming a subset of said plurality of factors, said subset of said plurality of factors including one or more factors, wherein an absolute value of a coefficient of correlation between a set of values assigned to any factor of said one or more factors and said plurality of activity levels is greater than or equal to said pre-defined threshold value; and
means for generating a regression model to predict an activity level, said regression model based on said plurality of activity levels and said subset of said plurality of factors.
-
-
17. (canceled)
-
18. (canceled)
-
19. A computer program product comprising a computer-usable medium including computer-usable program code for predicting a user activity level associated with an application executing on a computing system in a multi-user computing environment, said computer program product comprising:
-
computer-usable code for defining a user activity metric (UAM) as a variable, said UAM capable of being assigned a plurality of activity levels, wherein each activity level of said plurality of activity levels is a measured number of transactions performed by a plurality of users utilizing said application during a time period of a plurality of time periods, or a measured number of said users utilizing said application during said time period;
computer-usable code for defining a plurality of factors associated with said plurality of activity levels, said plurality of factors including a calendar type factor, said calendar type factor capable of being assigned a plurality of values of a cyclic time scale, wherein each value of said plurality of values represents an amount of time elapsed from a specified event;
computer-usable code for assigning said plurality of activity levels to said UAM;
computer-usable code for assigning said plurality of values to said calendar type factor of said plurality of factors, said plurality of values associated with said plurality of activity levels in a one-to-one correspondence, and with said plurality of time periods in a one-to-one correspondence;
computer-usable code for adjusting, subsequent to said assigning said plurality of values to said calendar type factor, said plurality of values to a plurality of adjusted values, said plurality of adjusted values including a pre-defined base value and one or more other adjusted values, wherein an adjusted value of said one or more other adjusted values represents a distance to said pre-defined base value, and wherein said adjusting includes facilitating a representation of a linear dependency between said plurality of adjusted values and said plurality of activity levels;
computer-usable code for calculating a plurality of coefficients of correlation, said calculating said plurality of coefficients of correlation including calculating a first coefficient of correlation between said plurality of adjusted values and said plurality of activity levels;
computer-usable code for determining that an absolute value of said first coefficient of correlation is less than a pre-defined threshold value;
computer-usable code for excluding, in response to said determining, said calendar type factor from said plurality of factors, said excluding including forming a subset of said plurality of factors, said subset including one or more factors, wherein an absolute value of a second coefficient of correlation between a set of values assigned to any factor of said one or more factors and said plurality of activity levels is greater than or equal to said pre-defined threshold value; and
computer-usable code for generating a regression model to predict an activity level, said regression model based on said plurality of activity levels and said subset. - View Dependent Claims (20, 21, 24)
-
-
22. (canceled)
-
23. (canceled)
-
25. A computer program product comprising a computer-usable medium including computer-usable program code for predicting a user activity level associated with an application executing on a computing system in a multi-user computing environment, said computer program product comprising:
-
computer-usable code for defining a user activity metric (UAM) as a variable, said UAM capable of being assigned a plurality of activity levels, wherein each activity level of said plurality of activity levels is a measured number of transactions performed by a plurality of users utilizing said application during a time period of a plurality of time periods, or a measured number of said plurality of users utilizing said application during said time period;
computer-usable code for defining a plurality of factors associated with said plurality of activity levels, said plurality of factors including a first subset of said plurality of factors, said first subset including at least one factor, wherein each factor of said first subset is a measurable variable on which said UAM depends, wherein said measurable variable is not a calendar type factor capable of being assigned amounts of time elapsed from a specified event;
computer-usable code for assigning said plurality of activity levels to said UAM, said computer-usable code for assigning said plurality of activity levels including computer-usable code for assigning a plurality of sets of at least two activity levels, wherein each set of said at least two activity levels depends upon a factor of said plurality of factors;
computer-usable code for assigning a plurality of values to each factor of said plurality of factors, said plurality of factors associated with said plurality of activity levels in a one-to-one correspondence, and with said plurality of time periods in a one-to-one correspondence;
computer-usable code for generating a first matrix of xij elements, wherein xij is a value of an i-th factor of said first subset of said plurality of factors, said value of said i-th factor measured at an end of a j-th time period of said plurality of time periods;
computer-usable code for generating a second matrix including said first matrix, wherein said second matrix is a three dimensional matrix including yijk elements, wherein yijk is an (i,j,k)-th activity level assigned to said UAM, and included in a k-th set of said plurality of sets of at least two activity levels, said (i,j,k)-th activity level associated with said xij and measured at an end of a (j+k)-th time period of said plurality of time periods;
computer-usable code for calculating a plurality of coefficients of correlation, wherein each coefficient of correlation of said plurality of coefficients of correlation is rik and wherein said rik is an (i,k)-th coefficient of correlation between said i-th factor of said first subset of said plurality of factors and said k-th set of said plurality of sets of at least two activity levels;
computer-usable code for determining that each absolute value of rik coefficients of correlation associated with said i-th factor is less than a pre-defined threshold value;
computer-usable code for excluding, in response to said determining, said i-th factor from said plurality of factors, said computer-usable code for excluding including computer-usable code for forming a subset of said plurality of factors, said subset of said plurality of factors including one or more factors wherein an absolute value of a coefficient of correlation between a set of values assigned to any factor of said one or more factors and said plurality of activity levels is greater than or equal to said pre-defined threshold value; and
computer-usable code for generating a regression model to predict an activity level, said regression model based on said plurality of activity levels and said subset of said plurality of factors.
-
-
26. (canceled)
-
27. (canceled)
-
28. A method for deploying computing infrastructure, comprising integrating computer-readable code into a computing system, wherein the code in combination with the computing system is capable of performing a process of predicting a user activity level associated with an application executing on a computing system in a multi-user computing environment, said process comprising:
-
defining a user activity metric (UAM) as a variable, said UAM capable of being assigned a plurality of activity levels, wherein each activity level of said plurality of activity levels is a measured number of transactions performed by a plurality of users utilizing said application during a time period of a plurality of time periods, or a measured number of said users utilizing said application during said time period;
defining a plurality of factors associated with said plurality of activity levels, said plurality of factors including a calendar type factor, said calendar type factor capable of being assigned a plurality of values of a cyclic time scale, wherein each value of said plurality of values represents an amount of time elapsed from a specified event;
assigning said plurality of activity levels to said UAM;
assigning said plurality of values to said calendar type factor of said plurality of factors, said plurality of values associated with said plurality of activity levels in a one-to-one correspondence, and with said plurality of time periods in a one-to-one correspondence;
adjusting, subsequent to said assigning said plurality of values to said calendar type factor, said plurality of values to a plurality of adjusted values, said plurality of adjusted values including a pre-defined base value and one or more other adjusted values, wherein an adjusted value of said one or more other adjusted values represents a distance to said pre-defined base value, and wherein said adjusting includes facilitating a representation of a linear dependency between said plurality of adjusted values and said plurality of activity levels;
calculating a plurality of coefficients of correlation, said calculating said plurality of coefficients of correlation including calculating a first coefficient of correlation between said plurality of adjusted values and said plurality of activity levels;
determining that an absolute value of said first coefficient of correlation is less than a pre-defined threshold value;
excluding, in response to said determining, said calendar type factor from said plurality of factors, said excluding including forming a subset of said plurality of factors, said subset including one or more factors, wherein an absolute value of a second coefficient of correlation between a set of values assigned to any factor of said one or more factors and said plurality of activity levels is greater than or equal to said pre-defined threshold value; and
generating a regression model to predict an activity level, said regression model based on said plurality of activity levels and said subset. - View Dependent Claims (29, 30, 33)
-
-
31. (canceled)
-
32. A method for deploying computing infrastructure, comprising integrating computer-readable code into a computing system, wherein the code in combination with the computing system is capable of performing a process of predicting a user activity level associated with an application executing on a computing system in a multi-user computing environment, said process comprising:
-
defining a user activity metric (UAM) as a variable, said UAM capable of being assigned a plurality of activity levels, wherein each activity level of said plurality of activity levels is a measured number of transactions performed by a plurality of users utilizing said application during a time period of a plurality of time periods, or a measured number of said users utilizing said application during said time period;
defining a plurality of factors associated with said UAM;
assigning a plurality of activity levels to said UAM;
assigning a plurality of values to a factor of said plurality of factors, said plurality of values associated with said plurality of activity levels in a one-to-one correspondence, and with said plurality of time periods in a one-to-one correspondence;
calculating at least one coefficient of correlation between at least two values of said plurality of values and at least two activity levels of said plurality of activity levels;
determining a maximum coefficient of correlation of said at least one coefficient of correlation that is less than a pre-defined threshold value;
excluding, in response to said determining, said factor from said plurality of factors to facilitate forming a subset of said plurality of factors, said subset including one or more factors, wherein an absolute value of a coefficient of correlation between any factor of said one or more factors and said UAM is greater than or equal to said pre-defined threshold value; and
generating a regression model to predict an activity level, said regression model based on said plurality of activity levels and said subset, wherein said defining said plurality of factors comprises defining said factor as a calendar type factor for which each value of said plurality of values represents an amount of time elapsed from a specified event, wherein said at least one coefficient of correlation comprises a single coefficient of correlation, and wherein said maximum coefficient of correlation is an absolute value of said single coefficient of correlation;
determining a sample mean, a standard deviation, and a standard error associated with said plurality of activity levels, said plurality of activity levels being historical data;
modifying a value of said plurality of values to a pre-defined base value, wherein said value is associated with an activity level of said plurality of activity levels exceeding said sample mean by more than said standard error multiplied by a pre-defined number; and
adjusting other values of said plurality of values to a set of two or more adjusted values, wherein an adjusted value of said two or more adjusted values represents a distance to said base value, wherein said calculating utilizes said base value and said set of two or more adjusted values to calculate said single coefficient of correlation, wherein a first absolute value of said single coefficient of correlation is greater than a second absolute value of another coefficient of correlation between said plurality of values and said plurality of activity level.
-
-
34. A method of predicting a user activity level associated with an application executing on a computing system in a multi-user computing environment, said method comprising:
-
defining a user activity metric (UAM) as a variable, said UAM capable of being assigned a plurality of activity levels, wherein each activity level of said plurality of activity levels is a measured number of transactions performed by a plurality of users utilizing said application during a time period of a plurality of time periods, or a measured number of said plurality of users utilizing said application during said time period;
defining a plurality of factors associated with said plurality of activity levels said plurality of factors including a first subset of said plurality of factors, said first subset including at least one factor, wherein each factor of said first subset is a measurable variable on which said UAM depends, wherein said measurable variable is not a calendar type factor capable of being assigned amounts of time elapsed from a specified event;
assigning said plurality of activity levels to said UAM, said assigning said plurality of activity levels including assigning a plurality of sets of at least two activity levels, wherein each set of said at least two activity levels depends upon a factor of said plurality of factors;
assigning a plurality of values to each factor of said plurality of factors, said plurality of factors associated with said plurality of activity levels in a one-to-one correspondence, and with said plurality of time periods in a one-to-one correspondence;
generating a first matrix of xij elements, wherein xij is a value of an i-th factor of said first subset of said plurality of factors, said value of said i-th factor measured at an end of a j-th time period of said plurality of time periods;
generating a second matrix including said first matrix, wherein said second matrix is a three dimensional matrix including yijk elements, wherein yijk is an (i,j,k)-th activity level assigned to said UAM, and included in a k-th set of said plurality of sets of at least two activity levels, said (i,j,k)-th activity level associated with said xij and measured at an end of a (j+k)-th time period of said plurality of time periods;
calculating a plurality of coefficients of correlation, wherein each coefficient of correlation of said plurality of coefficients of correlation is rik and wherein said rik is an (i,k)-th coefficient of correlation between said i-th factor of said first subset of said plurality of factors and said k-th set of said plurality of sets of at least two activity levels;
determining that each absolute value of rik coefficients of correlation associated with said i-th factor is less than a pre-defined threshold value;
excluding, in response to said determining, said i-th factor from said plurality of factors, said excluding including forming a subset of said plurality of factors, said subset of said plurality of factors including one or more factors, wherein an absolute value of a coefficient of correlation between a set of values assigned to any factor of said one or more factors and said plurality of activity levels is greater than or equal to said pre-defined threshold value; and
generating a regression model to predict an activity level, said regression model based on said plurality of activity levels and said subset of said plurality of factors.
-
-
35. A method for deploying computing infrastructure, comprising integrating computer-readable code into a computing system, wherein the code in combination with the computing system is capable of performing a process of predicting a user activity level associated with an application executing on a computing system in a multi-user computing environment, said process comprising:
-
defining a user activity metric (UAM) as a variable, said UAM capable of being assigned a plurality of activity levels, wherein each activity level of said plurality of activity levels is a measured number of transactions performed by a plurality of users utilizing said application during a time period of a plurality of time periods, or a measured number of said users utilizing said application during said time period;
defining a plurality of factors associated with said UAM;
assigning a plurality of activity levels to said UAM;
assigning a plurality of values to a factor of said plurality of factors, said plurality of values associated with said plurality of activity levels in a one-to-one correspondence, and with said plurality of time periods in a one-to-one correspondence;
calculating at least one coefficient of correlation between at least two values of said plurality of values and at least two activity levels of said plurality of activity levels;
determining a maximum coefficient of correlation of said at least one coefficient of correlation that is less than a pre-defined threshold value;
excluding, in response to said determining, said factor from said plurality of factors to facilitate forming a subset of said plurality of factors, said subset including one or more factors, wherein an absolute value of a coefficient of correlation between any factor of said one or more factors and said UAM is greater than or equal to said pre-defined threshold value; and
generating a regression model to predict an activity level, said regression model based on said plurality of activity levels and said subset, wherein said defining said plurality of factors comprises;
defining a first subset of said plurality of factors, said first subset including at least one factor, wherein said factor is included in said first subset, wherein each factor of said first subset is a measurable variable on which said UAM depends, wherein said measurable variable is not a calendar type factor for which each value of said plurality of values represents an amount of time elapsed from a specified event, and wherein said process further comprises;
assigning said plurality of activity levels to said UAM, said plurality of activity levels included in a plurality of sets of at least two activity levels, wherein each set of said at least two activity levels depends upon a factor of said plurality of factors, and is included in said plurality of activity levels, and wherein said plurality of activity levels is included in one set of said plurality of sets;
generating a first matrix of xij elements, wherein xij is a value of an i-th factor of said first subset measured at an end of a j-th time period of said plurality of time periods; and
generating a second matrix including said first matrix, wherein said second matrix is a three dimensional matrix including yijk elements, wherein yijk is an (i,j,k)-th activity level assigned to said UAM, and included in a k-th set of said plurality of sets of at least two activity levels, said (i,j,k)-th activity level associated with said xij and measured at an end of a (j+k)-th time period of said plurality of time periods, and wherein said second matrix includes elements rik, wherein rik is an (i,k)-th coefficient of correlation between said i-th factor of said first subset and said k-th set, wherein said maximum coefficient of correlation is an absolute value of an element (r) included in a set of said elements rik, wherein said absolute value of said r is greater than or equal to an absolute value of any other element of said set of said elements rik, said r being associated with said factor and with a set of said plurality of sets of at least two activity levels, and wherein said maximum coefficient of correlation is associated with a time step K, wherein each activity level of said set of said plurality of sets is measured K units of time after an end of a period of time of said plurality of periods of time. - View Dependent Claims (36)
-
Specification