Lithofacies analysis serves as the foundation for reservoir evaluation. However

due to the limited coring quantity and cost constraints

identifying lithofacies using logging data for uncored wells becomes crucial. In the Longdong area of the Ordos Basin

the lithofacies of the Chang-7 member have been classified into six types dependent on core identification results and imaging logging data. Based on core calibration

the logging response characteristics of different lithofacies were summarized

leading to the establishment of the lithofacies recognition mode using conventional logging curve. To achieve automatic lithofacies recognition in the study area

machine learning algorithms were employed. The traditional classification algorithms were affected significantly by the unbalanced sample. After comparing the application effects of different unbalanced data classification algorithm in the region

it’s found that bagging algorithm of ensemble learning notably improved the classification performance of all lithofacies by combining multiple classifiers. As a result

the overall lithofacies identification precision of this region has been improved by 20 %. According to the regional application results

the identification accuracy of single well can reach 84.33 %

demonstrating its practical applicability and effectiveness.