Parameter-dependent model-blending with multi-expert based machine learning and proxy sites

A system to classify sites to obtain a proxy site for a site of interest, the system comprising:

• a memory device configured to store an input of a number of types of sites, wherein the site of interest and each of the sites has a corresponding latitude, longitude, and elevation;
• and
• a processor configured to determine critical parameters among parameters estimated by one or more models associated with each of the sites, the parameters including the critical parameters, a parameter of interest, and other parameters, and the critical parameters being determined to be critical in estimation of the parameter of interest according to the one or more models, to group two or more sites together in a same group when the two or more sites have same critical parameters, to classify the two or more sites of each group into a type, the number of types of sites being specified in the input, wherein the processor configured to classify the two or more sites in each group type is further configured to associate each of the two or more sites with a set of polynomial coefficients, to receive an indication of a number of the types and to classify each of the two or more sites of each group into a type using an unsupervised clustering machine learning algorithm based on the received number of the types and based on the set of polynomial coefficients associated with each of the two or more sites of each group, wherein the processor configured to classify is further configured to obtain the set of polynomial coefficients associated with each of the two or more sites of each group by fitting polynomial models to first and second order dependence of error in estimation of the parameter of interest on each of the critical parameters associated with each of the two or more sites and wherein the processor configured to obtain the set of polynomial coefficients is further configured to examine the first order dependence by translating a first order error dependence curve to generate two polynomial coefficients and to examine the second order dependence by translating a second order error dependence surface into six polynomial coefficients, to correlate the type associated with each site with the latitude, the longitude, and the elevation of the site, wherein the processor correlates the type associated with each site with latitude, longitude, and elevation of the site based on training a classification machine learning model to correlate the site type with the latitude, the longitude, and the elevation of the site, to obtain a proxy site for the site of interest from among the sites by determining the type correlated with the longitude, the latitude, and the elevation of the site of interest, and to obtain training data for use with one or more machine learning algorithms using data obtained at the proxy site based on the data from the site of interest being unavailable.

A non-transitory computer program product having computer readable instructions stored thereon which, when executed by a processor, cause the processor to implement a method of classifying sites to obtain a proxy site for a site of interest, the method comprising:

• determining critical parameters among parameters estimated by one or more models associated with each of the sites, the parameters including the critical parameters, a parameter of interest, and other parameters, and the critical parameters being determined to be critical in estimation of the parameter of interest according to the one or more models, wherein the site of interest and each of the sites has a corresponding latitude, longitude, and elevation;
• grouping two or more sites together in a same group when the two or more sites have same critical parameters;
• for each group of the two or more sites, classifying the two or more sites by type, wherein the classifying the two or more sites in each group by type includes associating each of the two or more sites with a set of polynomial coefficients, receiving an indication of a number of the types and classifying each of the two or more sites of each group into a type using an unsupervised clustering machine learning algorithm based on the received number of the types and based on the set of polynomial coefficients associated with each of the two or more sites of each group, wherein the classifying further comprises obtaining the set of polynomial coefficients associated with each of the two or more sites of each group by fitting polynomial models to first and second order dependence of error in estimation of the parameter of interest on each of the critical parameters associated with each of the two or more sites and wherein obtaining the set of polynomial coefficients further comprises examining the first order dependence by translating a first order error dependence curve to generate two polynomial coefficients and examining the second order dependence by translating a second order error dependence surface into six polynomial coefficients;
• correlating the type associated with each site with the latitude, the longitude, and the elevation of the site, wherein the correlating includes training the classification machine learning model to correlate the site type with the latitude, the longitude, and the elevation of the site;
• obtaining the proxy site for the site of interest from among the sites by determining the type correlated with the latitude, the longitude, and the elevation of the site of interest;
• and
• obtaining training data for use with one or more machine learning algorithms using data obtained at the proxy site based on the data from the site of interest being unavailable.