![Scientific and technical journal «Priborostroenie»](/images/mag-pr.png)
SEARCH FOR ANALOGUE DEPOSITS BASED ON BAYESIAN NETWORKS CLUSTERING
![Scientific and technical journal «Priborostroenie»](/images/mag-pr.png)
Annotation
An algorithm is developed to search for oil and gas deposits-analogues, based on clustering of Bayesian networks, built on parameters of the known deposits. Using Bayesian networks, it is possible to effectively represent oil and gas fields in the form of multivariate distributions, accounting for the complex relationships between the parameters. For each of the deposits in the database, a Bayesian network was built on a sample of its nearest neighbors obtained using the cosine distance metric. Clustering of the Bayesian networks built on the samples is performed by comparing the Hamming distance metric between adjacency matrices stretched into a one-dimensional vector. The developed approach is compared to other analogue search methods based on machine learning. Results of evaluation of the algorithm performance are presented, confirming that modeling and searching for analogues using Bayesian networks is a more comprehensive solution to the problem. The accuracy of restoring missing values for most parameters using the developed algorithm turned out to be higher than in existing classical clustering algorithms.
Keywords
Постоянный URL
Articles in current issue
- APPLICATION OF POLYNOMIAL CODES IN THE SYNTHESIS OF BUILT-IN CONTROL CIRCUITS FOR COMBINATIONAL DEVICES USING BOOLEAN COMPLEMENT METHOD
- SYNTHESIS OF DISCRETE BANDPASS AND NOTCH FILTERS USING INVARIANT IMPULSE AND TRANSIENT CHARACTERISTICS
- PREFERRED PAIRS OF GMW SEQUENCES WITH PERIOD N=1023 FOR DIGITAL INFORMATION TRANSMISSION SYSTEMS
- PRECISION ELECTRIC DRIVE BASED ON A MULTI-PHASE SYNCHRONOUS MOTOR
- INCREASING THE ACCURACY AND SPEED OF THE OPTICAL RANGEFINDER DUE TO CORRELATION SIGNAL PROCESSING
- MODERNIZATION OF THE LASER HEAD HOUSING DESIGN BASED ON THE CAPABILITIES OF ADDITIVE EQUIPMENT
- METHOD FOR EVALUATING THE LINEAR AND VOLUMETRIC WEAR RESISTANCE OF MATERIALS