DICTIONARY REDUCTION TECHNIQUE FOR BUSINESS NAME CATEGORIZATION
An energy management system includes a neural network, a predictive model, and a dictionary reducer. The network iteratively calculates weights, resulting in a final set, for each of a plurality of single-word terms and word order terms taken from training data business names, where each of the weights is indicative of a likelihood of correlating one of a plurality of business categories. The predictive employs sets of the weights to predict a first corresponding one of the plurality of business categories for each of the training data business names until employment of the final set accurately predicts a correct business category for the each of the training data business names, and subsequently employs the final set of the weights to predict a second corresponding one of the plurality of business categories for each of a plurality of operational business names. The dictionary reducer eliminates unessential terms taken to determine the plurality of single-word terms and word order terms.
1 . An energy management system, for communicating with one or more buildings for purposes of managing energy consumption of devices within the buildings, the energy management system comprising:
a neural network, configured to iteratively calculate weights, resulting in a final set of said weights, for each of a plurality of single-word terms and word order terms taken from training data business names, wherein each of said weights is indicative of a likelihood of correlating one of a plurality of business categories, and wherein said word order terms comprise each of said single-word terms along with a corresponding order in said training data business names;
a predictive model, coupled to said neural network, configured to iteratively employ sets of said weights to predict a first corresponding one of said plurality of business categories for each of said training data business names until employment of said final set of said weights accurately predicts a correct business category for said each of said training data business names, and configured to subsequently employ said final set of said weights to predict a second corresponding one of said plurality of business categories for each of a plurality of operational business names; and
a dictionary reducer, coupled to said neural network, configured to eliminate unessential single-word terms and word order terms taken from said training data business names to determine said plurality of single-word terms and word order terms, wherein said plurality of single-word terms and word order terms are essential to predicting said correct business category for said each of said training data business names.
2 . The energy management system as recited in claim 1 , wherein if one set of said weights does not provide for accurate prediction of a correct business category for said each of said training data business names, then said dictionary reducer includes formerly eliminated single-word terms and word order terms to determine said plurality of single-word terms and word order terms until accurate prediction of a correct business category for said each of said training data business names is achieved.
3 . The energy management system as recited in claim 1 , wherein said dictionary reducer eliminates said unessential single-word terms and word order terms taken from said training data business names to determine said plurality of single-word terms and word order terms as a function of a performance metric corresponding to a vertical categorizer within which said neural network, said predictive model, and said dictionary reducer are disposed.
4 . The energy management system as recited in claim 1 , wherein said dictionary reducer generates score vectors for all single-word terms and word order terms taken from said training data business, wherein a given score vector indicates a degree to which removal of a corresponding single-word term affects determination of said correct business category.
5 . The energy management system as recited in claim 1 , wherein said second corresponding one of said plurality of business categories for each of a plurality of operational business names is employed within a network operations center (NOC) to determine and transmit to the buildings messages appropriate for managing energy consumption of devices within the buildings.
6 . The energy management system as recited in claim 5 , wherein said messages are transmitted over physical communication links to physical nodes within some of the buildings.
7 . The energy management system as recited in claim 5 , wherein said messages are transmitted over virtual communication links to virtual nodes within some of the buildings.
8 . An energy management system, for communicating with one or more buildings for purposes of managing energy consumption of devices within the buildings, the energy management system comprising:
a network operations center (NOC), coupled to each of the buildings via one or more communication channels, said NOC comprising:
a neural network, configured to iteratively calculate weights, resulting in a final set of said weights, for each of a plurality of single-word terms and word order terms taken from training data business names, wherein each of said weights is indicative of a likelihood of correlating one of a plurality of business categories, and wherein said word order terms comprise each of said single-word terms along with a corresponding order in said training data business names;
a predictive model, coupled to said neural network, configured to iteratively employ sets of said weights to predict a first corresponding one of said plurality of business categories for each of said training data business names until employment of said final set of said weights accurately predicts a correct business category for said each of said training data business names, and configured to subsequently employ said final set of said weights to predict a second corresponding one of said plurality of business categories for each of a plurality of operational business names; and
a dictionary reducer, coupled to said neural network, configured to eliminate unessential single-word terms and word order terms taken from said training data business names to determine said plurality of single-word terms and word order terms, wherein said plurality of single-word terms and word order terms are essential to predicting said correct business category for said each of said training data business names.
9 . The energy management system as recited in claim 8 , wherein if one set of said weights does not provide for accurate prediction of a correct business category for said each of said training data business names, then said dictionary reducer includes formerly eliminated single-word terms and word order terms to determine said plurality of single-word terms and word order terms until accurate prediction of a correct business category for said each of said training data business names is achieved.
10 . The energy management system as recited in claim 8 , wherein said dictionary reducer eliminates said unessential single-word terms and word order terms taken from said training data business names to determine said plurality of single-word terms and word order terms as a function of a performance metric corresponding to a vertical categorizer within which said neural network, said predictive model, and said dictionary reducer are disposed.
11 . The energy management system as recited in claim 8 , wherein said dictionary reducer generates score vectors for all single-word terms and word order terms taken from said training data business, wherein a given score vector indicates a degree to which removal of a corresponding single-word term affects determination of said correct business category.
12 . The energy management system as recited in claim 8 , wherein said second corresponding one of said plurality of business categories for each of a plurality of operational business names is employed by said NOC to determine and transmit to the buildings messages appropriate for managing energy consumption of devices within the buildings.
13 . The energy management system as recited in claim 12 , wherein said communication links comprise physical communication links, and wherein said messages are transmitted over said physical communication links to physical nodes within some of the buildings.
14 . The energy management system as recited in claim 12 , wherein said communication links comprise virtual communication links, and wherein said messages are transmitted over said virtual communication links to virtual nodes within some of the buildings.
15 . A method for communicating with one or more buildings for purposes of managing energy consumption of devices within the buildings, the method comprising:
via a neural network, iteratively calculating weights, resulting in a final set of the weights, for each of a plurality of single-word terms and word order terms taken from training data business names, wherein each of the weights is indicative of a likelihood of correlating one of a plurality of business categories, and wherein said word order terms comprise each of said single-word terms along with a corresponding order in said training data business names;
via a predictive model coupled to the neural network, iteratively employing sets of the weights to predict a first corresponding one of the plurality of business categories for each of the training data business names until employment of the final set of the weights accurately predicts a correct business category for the each of the training data business names, and subsequently employing the final set of the weights to predict a second corresponding one of the plurality of business categories for each of a plurality of operational business names; and
via a dictionary reducer coupled to the neural network, eliminating unessential single-word terms and word order terms taken from the training data business names to determine the plurality of single-word terms and word order terms, wherein the plurality of single-word terms and word order terms is essential to predicting the correct business category for the each of the training data business names.
16 . The method as recited in claim 15 , wherein if one set of the weights does not provide for accurate prediction of a correct business category for the each of the training data business names, then the dictionary reducer includes formerly eliminated single-word terms and word order terms to determine the plurality of single-word terms and word order terms until accurate prediction of a correct business category for the each of the training data business names is achieved.
17 . The method as recited in claim 15 , wherein the dictionary reducer eliminates the unessential single-word terms and word order terms taken from the training data business names to determine the plurality of single-word terms and word order terms as a function of a performance metric corresponding to a vertical categorizer within which the neural network, the predictive model, and the dictionary reducer are disposed.
18 . The method as recited in claim 15 , wherein the dictionary reducer generates score vectors for all single-word terms and word order terms taken from the training data business, wherein a given score vector indicates a degree to which removal of a corresponding single-word term affects determination of the correct business category.
19 . The method as recited in claim 15 , wherein the second corresponding one of the plurality of business categories for each of a plurality of operational business names is employed within a network operations center (NOC) to determine and transmit to the buildings messages appropriate for managing energy consumption of devices within the buildings.
20 . The method as recited in claim 19 , wherein the messages are transmitted over physical communication links to physical nodes within some of the buildings.
21 . The method as recited in claim 19 , wherein the messages are transmitted over virtual communication links to virtual nodes within some of the buildings.