The integrity of geomagnetic data is a critical factor for understanding the evolutionary process of Earth's magnetic field, as it can provide useful information for near-surface exploration, unexploded explosive ordnance (UXO) detection, etc. Aimed to reconstruct geomagnetic data from under-sampled or missing traces, this paper presented an approach based on machine learning techniques to avoid the time & labor-intensive nature of the traditional manual and linear interpolation approaches. In this study, three classic machine learning models, support vector machine (SVM), random forests and gradient boosting were built. The proposed learning models were first used to specify a continuous regression hyperplane from training data, to recognize the probably intrinsic relation between missing and completed traces. Afterwards, the trained models were used to reconstruct the missing geomagnetic traces for validation, while testing other new field data. Finally, numerical experiments were derived. The results showed that the performance of our methods was more competitive in comparison with the traditional linear method, as the reconstruction accuracy was increased by approximately 10% ～ 15%.
展开▼
