Predicting the remaining time before the next earthquake based on seismic signals generated in a laboratory setting is a challenging research task that is of significant importance for earthquake hazard assessment. In this study, we employed a mel spectrogram and the mel frequency cepstral coefficient (MFCC) to extract relevant features from seismic signals. Furthermore, we proposed a deep learning model with a hierarchical structure. This model combines the characteristics of long short-term memory (LSTM), one-dimensional convolutional neural networks (1D-CNN), and two-dimensional convolutional neural networks (2D-CNN). Additionally, we applied a stacking model fusion strategy, combining gradient boosting trees with deep learning models to achieve optimal performance. We compared the performance of the aforementioned feature extraction methods and related models for earthquake prediction. The results revealed a significant improvement in predictive performance when the mel spectrogram and stacking were introduced. Additionally, we found that the combination of 1D-CNN and 2D-CNN has unique advantages in handling time-series problems.

Multivariate time-series anomaly detection is important for server monitoring, industrial systems, and intelligent operation and maintenance. Existing point-wise and point-adjusted metrics often fail to show whether an anomalous segment is detected early enough for intervention. This paper proposes a lightweight dual-score enhancement for Anomaly Transformer. The method retains the original multiplicative anomaly score and adds window-normalized reconstruction error and association discrepancy terms, so weak devi- ations near anomaly onsets become more visible without changing the backbone network or using additional labels. Experiments are conducted on four public multivariate bench- marks, namelySMD,PSM,SMAP,andMSL,underthreerandomseeds. Comparedwiththe original Anomaly Transformer score, the proposed early-warning union mode improves Raw Event Recall by 0.3435, 0.2408, 0.0995, and 0.0278 on SMD, PSM, SMAP, and MSL, respectively, and reduces mean detection delay by 53.66%, 49.97%, 52.69%, and 39.74%. Additional range- and proximity-aware metrics further show that the score-level enhance- ment improves segment coverage and early-warning behavior, while AUPRC remains dataset-dependent. The results indicate that association-discrepancy score enhancement is a practical complement to existing Transformer-based anomaly detectors when early warning is more important than point-wise matching alone.