Insufficient Sampling Data Research Articles

An instance-based transfer learning model with attention mechanism for freight train travel time prediction in the China–Europe railway express

Since the inception of the China–Europe Railway Express (CRE), rail transportation has emerged as a predominant means of transporting goods across international borders. The quality and reliability of railway transport services are constrained by the precision of train travel times. However, the real-time data available for the CRE trains is rather limited due to the volume and duration of operations, which increases the difficulty of forecasting CRE train travel times. In order to enhance the accuracy of predictions, a novel two-stage transfer learning model with the Long Short-Term Memory attention (T.R2_LSTM_A) is introduced in this study. By incorporating attention mechanisms, the Long Short-Term Memory networks (LSTM) module, the two-stage TrAdaBoost.R2 algorithm, and transfer learning (TL), the model captures the inherent characteristics of time series data and attains a high level of accuracy in predictions. This approach effectively mitigates the limitations of current transfer learning methods, which are primarily caused by insufficient sample data. In the model, an instance-based TL method is utilized to address the shortcomings of insufficient data, and an attention-based LSTM is selected as a base learner to overcome the thorny issues of gradient disappearance and explosion. Additionally, LSTM Attention (LSTM_A), a novel attention mechanism, is introduced to improve the ability of the proposed model to capture complicated characteristics. The T.R2_LSTM_A model is systemically evaluated and shown to be superior to Random Forest (RF), Support Vector Regression (SVR), LSTM, convolution-based LSTM (CLSTM), AdaBoost_LSTM, LSTM with attention mechanism (LSTM_Attention), and TrAdaBoost.R2_RF (T.R2_RF) for predicting train travel time, using a real-life case study of CRE. The MSE, RMSE, MAPE, and MAE values for the T.R2_LSTM_A model are equal to 5.572, 2.361, 2.578, and 1.767, respectively. These values surpass those of other widely used prediction methods, indicating that the T.R2_LSTM_A model is well-suited for predicting train travel times in the CRE.

On Integrating Time-Series Modeling with Long Short-Term Memory and Bayesian Optimization: A Comparative Analysis for Photovoltaic Power Forecasting

The means of energy generation are rapidly progressing as production shifts from a centralized model to a fully decentralized one that relies on renewable energy sources. Energy generation is intermittent and difficult to control owing to the high variability in the weather parameters. Consequently, accurate forecasting has gained increased significance in ensuring a balance between energy supply and demand with maximum efficiency and sustainability. Despite numerous studies on this issue, large sample datasets and measurements of meteorological variables at plant sites are generally required to obtain a higher prediction accuracy. In practical applications, we often encounter the problem of insufficient sample data, which makes it challenging to accurately forecast energy production with limited data. The Holt–Winters exponential smoothing method is a statistical tool that is frequently employed to forecast periodic series, owing to its low demand for training data and high forecasting accuracy. However, this model has limitations, particularly when handling time-series analysis for long-horizon predictions. To overcome this shortcoming, this study proposes an integrated approach that combines the Holt–Winters exponential smoothing method with long short-term memory and Bayesian optimization to handle long-range dependencies. For illustrative purposes, this new method is applied to forecast rooftop photovoltaic production in a real-world case study, where it is assumed that measurements of meteorological variables (such as solar irradiance and temperature) at the plant site are not available. Through our analysis, we found that by utilizing these methods in combination, we can develop more accurate and reliable forecasting models that can inform decision-making and resource management in this field.

Deep Learning-Based Technique for Remote Sensing Image Enhancement Using Multiscale Feature Fusion.

The present study proposes a novel deep-learning model for remote sensing image enhancement. It maintains image details while enhancing brightness in the feature extraction module. An improved hierarchical model named Global Spatial Attention Network (GSA-Net), based on U-Net for image enhancement, is proposed to improve the model's performance. To circumvent the issue of insufficient sample data, gamma correction is applied to create low-light images, which are then used as training examples. A loss function is constructed using the Structural Similarity (SSIM) and Peak Signal-to-Noise Ratio (PSNR) indices. The GSA-Net network and loss function are utilized to restore images obtained via low-light remote sensing. This proposed method was tested on the Northwestern Polytechnical University Very-High-Resolution 10 (NWPU VHR-10) dataset, and its overall superiority was demonstrated in comparison with other state-of-the-art algorithms using various objective assessment indicators, such as PSNR, SSIM, and Learned Perceptual Image Patch Similarity (LPIPS). Furthermore, in high-level visual tasks such as object detection, this novel method provides better remote sensing images with distinct details and higher contrast than the competing methods.

Comparison of variable extraction methods using surface field data and its key influencing factors: A case study on aboveground biomass of Pinus densata forest using the original bands and vegetation indices of Landsat 8

Insufficient sample data is a challenge when estimating forest aboveground biomass (AGB) using large-scale remote sensing. Extracting remote sensing information from sub-compartments could rectify such defects, but the corresponding method, its accuracy, and influencing factors still need to be clarified. We combined Landsat 8 data with a Pinus densata forest sub-compartment to extract remote sensing information that matched the sample plots. Six sub-compartment based methods, including the centroid point extraction method, and the minimum, mean, maximum, majority, and median statistic extraction methods were used to extract sub-compartment remote sensing information and compare the differences between each method and the true values. For each method, structural equation modelling (SEM) was used to explore the effect of sub-compartment topography, shape, and forest stand factors on the extraction error. Mean statistic was the best extraction method, with the highest consistency index, and the lowest mean relative error, between the extracted and true values. All three factors affected extraction accuracy, with forest stand being the dominant one. When sub-compartment data are sufficient, but sample plots are insufficient, it is an effective extrapolation method for large-scale AGB estimation.

A tensor based price evaluation approach for the used mobile phone recycling

The automatic price evaluation is an important and challenging issue for the used mobile phone recycling. However, due to complicated relationships between attributes and price as well as insufficient sample data, current approaches cannot achieve accurate price evaluation for the used mobile phones, which significantly reduce the efficiency for the used mobile phone recycling. To this end, a tensor based approach is proposed in this paper, which can achieve accurate and efficient price evaluation for used mobile phones. In the proposed approach, first, the mutual information based attributes selection mechanism and the boxplot based sample selection mechanism are employed to select the most relevant attributes and suitable price samples from used mobile phone dataset. Then, a tensor model is constructed to establish multi-dimensional relationships between selected attributes and prices of used mobile phones. Finally, the missing values in the constructed tensor model is completed through the gradient descent and CANDECOMP/PARAFAC decomposition algorithms. The completed tensor model can be directly used for the price evaluation of used mobile phones based on their attributes without further calculation. Based on the real price dataset of used mobile phones, the experiments indicate that the proposed approach outperforms most of current approaches for the price evaluation for the used mobile phones.

Calculation and analysis of agricultural carbon emission efficiency considering water–energy–food pressure: Modeling and application

Agricultural carbon emission efficiency (ACEE) measurement, a tool for effectively achieving sustainable development goals, has garnered much attention. However, the impacts of resource pressures such as water, energy and food on ACEE have been overlooked, and the high dimensionality of the measurement model and insufficient sample data can easily distort the measurement results. Therefore, from a green development perspective, we established a new ACEE measurement framework considering the water–energy–food pressure index and a new integrated ACEE measurement model (CSMA-PPE-USSBM) that includes chaotic maps, the slime mould algorithm (SMA), projection pursuit evaluation (PPE) and the undesirable super slack-based measure (USSBM). The model was used to calculate ACEE in 13 prefecture-level municipalities in Heilongjiang Province, China, and analyze its spatiotemporal evolution and possible causes. The results showed that this model avoids the above problems. The reliability coefficient and stability coefficient reached 0.962 and 0.971, respectively; ACEE in Heilongjiang Province has much room for improvement, but there are obvious differences in carbon emission efficiency in different carbon emission type regions. The key driving forces of ACEE variation can generate significant scale effects. Provincial driving factors can affect ACEE variation in prefecture municipalities, where the influence range is limited or the influence of driving factors gradually emerges. The research results provide a theoretical reference for accurately measuring the regional ACEE and analyzing the driving mechanism of ACEE and green agricultural development.

Multi-Object Detection for Inland Ship Situation Awareness Based on Few-Shot Learning

With the rapid development of artificial intelligence technology and unmanned surface vehicle (USV) technology, object detection and tracking have wide applications in marine monitoring and intelligent ships. However, object detection and tracking tasks on small sample datasets often face challenges due to insufficient sample data. In this paper, we propose a ship detection and tracking model with high accuracy based on a few training samples with supervised information based on the few-shot learning framework. The transfer learning strategy is designed, innovatively using an open dataset of vehicles on highways to improve object detection accuracy for inland ships. The Shuffle Attention mechanism and smaller anchor boxes are introduced in the object detection network to improve the detection accuracy of different targets in different scenes. Compared with existing methods, the proposed method is characterized by fast training speed and high accuracy with small datasets, achieving 84.9% (mAP@0.5) with only 585 training images.

Improved Prediction of Aquatic Beetle Diversity in a Stagnant Pool by a One-Dimensional Convolutional Neural Network Using Variational Autoencoder Generative Adversarial Network-Generated Data

Building a reasonable model for predicting biodiversity using limited data is challenging. Expanding limited experimental data using a variational autoencoder generative adversarial network (VAEGAN) to improve biodiversity predictions for a region is a new strategy. Aquatic beetle diversity in a large >30-year-old artificial pool that had not had human interference in Nanshe Village (Dapeng Peninsula, Shenzhen City, Guangdong Province, China) was investigated. Eight ecological factors were considered. These were water temperature, salinity, pH, water depth, proportional area of aquatic plants, proportional area of submerged plants, water area, and water level. Field sampling was performed for 1 or 2 days in the middle or late part of each month for a year. A type D net was swept 10 times in the same direction in each ~1 m × ~1 m sample square, generating 132 datasets (experimental data). In total, 39 aquatic beetle species were collected, 19 of which were assigned to Hydrophilidae, 16 to Dytiscidae, 3 to Noteridae, and 1 to Gyrinidae. A one-dimensional convolutional neural network (1-D CNN) was used to assess and predict the grade of the number of individuals and the number of aquatic beetle species. The Bayesian-optimized 1-D CNN established using 112 experimental datasets as the training set and the other 20 datasets as validation and testing sets gave a 74.0% prediction accuracy for the grade of the number of individuals and a 70.0% prediction accuracy for the number of species. The impact of insufficient sample data on the model was assessed using a VAEGAN to expand the training set from 112 to 512 samples, and then the Bayesian-optimized 1-D CNN-based VAEGAN prediction model was re-established. This improved prediction accuracy for the grade of the number of individuals to 86.0% and for the number of species to 85.0%. The grade of the number of individuals’ prediction accuracy was 88.0% and the number of species’ prediction accuracy was 85.0% when the random effects of only obtaining a single individual of a species were excluded. The results indicated that the accuracy of the 1-D CNN in predicting the aquatic beetle species number and abundance from relevant environmental factors can be improved using a VAEGAN to expand the experimental data.

A data-driven workflow based on multisource transfer machine learning for gas-bearing probability distribution prediction: A case study

Machine learning (ML) plays an important role in gas-bearing prediction based on multicomponent seismic data because it can reveal the complex relationship between reservoirs and their seismic attributes by establishing a nonlinear model. However, the degree of exploration in several areas remains limited, leading to insufficient sample data. Transfer learning (TL) can provide a feasible solution to this limitation by transferring learning information between data-rich sources and data-sparse target sources to accomplish a target task. Therefore, a data-driven workflow based on multisource TL is developed for predicting gas-bearing probability distribution to improve its accuracy in areas with limited sample data. First, the samples and labels of the synthetic data are established, and ML models (deep fully connected neural network and least-squares support vector machine) are applied to the synthetic data to predict the gas-bearing distribution. Subsequently, the samples and labels of the real data are established and completely mixed with the synthetic data to obtain the data set after TL. Finally, a supervised learning model is applied to predict the gas-bearing probability distribution in the study area. Adaptive mutation particle swarm optimization (AMPSO) is used to optimize the prediction performances of the two ML models. The prediction results of the AMPSO-ML are in good agreement with the actual drilling results. By comparing the prediction results after TL with those before TL, it is observed that the two models had higher efficiency and accuracy after TL, verifying the reliability of our scheme for predicting gas-bearing distribution. This method may be used for the effective prediction of gas-bearing probability distributions.

Reliability analysis of uncertain random systems based on uncertain differential equation

Typical failure modes of degradation and shock processes have been applied which are named as soft failures and hard failures, respectively. This paper discusses the reliability of uncertain random systems where randomness is regarded as objective indeterminacy with enough sample data, and uncertainty is referred to epistemic indeterminacy with insufficient sample data. Considering that the combination of internal and external factors, the degradation process is modelled by an uncertain differential equation, and external shocks are driven by an uncertain random renewal process where state variables are uncertain random variables. Chance measure is applied to define the reliability index which is distinguished from traditional reliability assessment methods that utilize probability measure. Three types of reliability models possessed with independent failures are presented, where the general reliability index formulas are provided. The analysis results of Caohai Nature Reserve show that uncertain random modelling method provides an effective approach to system reliability assessment.

Slope displacement prediction based on multisource domain transfer learning for insufficient sample data

Slope displacement prediction based on multisource domain transfer learning for insufficient sample data

Two-stage vibration-suppression framework for optimal robust placements design and reliable PID gains design via set-crossing theory and artificial neural network

• Atwo-stage controlmodelfor design of placements and PID parametersis proposed. • The optimal placementsrobust design of uncertain vibration systemsis proposed. • The optimal PID gainsdesign withtime-dependent reliabilitylimitsis proposed. • The ANNis appliedin vibratory controloptimization to improve the efficiency. Structure vibration control is a significant problem in practical engineering, and Proportional-Integral-Differential (abbreviated as “PID”) control is widely used. The design of a PID controller contains the robust design of placements of actuators and sensors (abbreviated as “ASPD”), and the reliable design of PID gains. Because of the uncertain parameters in practical engineering, the controller designed from the deterministic system has the possibility of failure when applied to a practical system. To solve this problem, this paper proposed a two-stage framework for uncertain vibration active control systems via non-probabilistic time-dependent reliability (abbreviated as “NTDR”) and artificial neutral network (abbreviated as “ANN”). This framework decouples the ASPD and PID gains design as two stages. Considering the insufficient sample data, uncertain parameters are quantified with non-probabilistic interval forms. The optimization objective of ASPD combines the deterministic eigenvalues of controllability gramian or observability gramian, the robust theory and placements constraints to carry out the optimization. In this way the optimization objective can be less affected by uncertain parameters and has practical significance. For the reliable design of PID gains, the NTDR is adapted to evaluate the safety of an uncertain system. Then ANNs from uncertain parameters and PID gains to the optimization objective considering NTDR are established. By carrying out the optimization to the last ANN, the efficiency of optimization is improved while the accuracy is guaranteed. Two numerical examples and one experimental example show the practicability and effectiveness of the proposed method.

Centralized contrastive loss with weakly supervised progressive feature extraction for fine-grained common thorax disease retrieval in chest x-ray.

Medical images have already become an essential tool for the diagnosis of many diseases. Thus a large number of medical images are being generated due to the daily routine inspection. An efficient image-based disease retrieval system will not only make full use of existing data, but also help physicians to prognosis the diseases. Medical image retrieval is represented by the classification and localization of common thorax diseases in x-ray images. Although extensive efforts have been put into this field, there are still manychallenges. Most of the existing fine-grained image research methods just apply existing deep learning frameworks in extracting the image features. However, these high-level features mainly focus on the global representations of the object, rather than simultaneously considering the local ones. It requires fine-grained details to classify the images with similar lesion areas. Thus, it is necessary to combine the global features and local ones to make the features more discriminative. On the other hand, training CNN models based on current existing strategies have a high time complexity, and is hard to get the discriminative features mentioned above. In addition, the visual retrieval method of fine-grained medical images still has the problem of insufficient sample data with accurate annotationinformation. To address above challenges, we introduced a novel fine-grained medical images retrieval method. First, a centralized contrastive loss (CCLoss) is proposed as our metric learning loss function. Parameters are updated by using the center point, which not only improves the distinguishing performance of features, but also effectively reduces the time complexity of the algorithm. In addition, a weakly supervised progressive feature extraction method is proposed to gradually extract the combined features. And the attention mechanism module is applied to screen the target information after the initial positioning for fine refinement, so as to separate the features with a high degree of discrimination. The retrieval of 14 different chest diseases is evaluated on the chest x-raydatasets. Compared with the existing research methods, the proposed method shows a better retrieval result for Recall@8 by 2.26 and achieves a very efficient training speed which is 100 times faster than the pair-wise loss-based training strategy. We also assessed the effects of Recall@k (k = 2, 4, 6, 8) for progressive features extracted from different steps to obtain a model with the best retrievalperformance. The proposed model is capable of learning discriminative representations from chest x-ray datasets, and it achieves better performance compared with other state-of-the-art methods. Therefore, the developed model would be useful in the diagnosis of common thorax disease or unknown chestdisease.

A GAN–SVR Prediction Method of the Metal Tube‐Bending Rebound with Small Samples

This paper investigates a predictive algorithm for the angle of the metal tube‐bending rebound with small samples. First, the generative adversarial network (GAN) approach is introduced to address the issues of insufficient sample data. The proposed method can realize data augmentation through a generator, enhancing training effectiveness compared to conventional model‐based and experimental prediction methods. To further reduce the problems caused by the small samples, the Wasserstein distance is utilized as the optimization objective for the GAN approach. Second, after obtaining the augmented dataset, Support Vector Regression (SVR) is employed to predict the rebound model of the metal tube‐bending. A novel predictive algorithm for the angle of the metal tube‐bending rebound based on GAN–SVR is proposed. It exhibits that the GAN–SVR owns more positive prediction ability and error when dealing with small samples than conventional GAN‐radial basis function methd (GAN–BP) and GAN–convolutional neural networks. Finally, the effectiveness of the proposed method is validated through experimental results.

Attack Behavior Extraction Based on Heterogeneous Cyberthreat Intelligence and Graph Convolutional Networks

The continuous improvement of the cyber threat intelligence sharing mechanism provides new ideas to deal with Advanced Persistent Threats (APT). Extracting attack behaviors, i.e., Tactics, Techniques, Procedures (TTP) from Cyber Threat Intelligence (CTI) can facilitate APT actors’ profiling for an immediate response. However, it is difficult for traditional manual methods to analyze attack behaviors from cyber threat intelligence due to its heterogeneous nature. Based on the Adversarial Tactics, Techniques and Common Knowledge (ATT&CK) of threat behavior description, this paper proposes a threat behavioral knowledge extraction framework that integrates Heterogeneous Text Network (HTN) and Graph Convolutional Network (GCN) to solve this issue. It leverages the hierarchical correlation relationships of attack techniques and tactics in the ATT&CK to construct a text network of heterogeneous cyber threat intelligence. With the help of the Bidirectional Encoder Representation from Transformers (BERT) pretraining model to analyze the contextual semantics of cyber threat intelligence, the task of threat behavior identification is transformed into a text classification task, which automatically extracts attack behavior in CTI, then identifies the malware and advanced threat actors. The experimental results show that F1 achieve 94.86% and 92.15% for the multi-label classification tasks of tactics and techniques. Extend the experiment to verify the method’s effectiveness in identifying the malware and threat actors in APT attacks. The F1 for malware and advanced threat actors identification task reached 98.45% and 99.48%, which are better than the benchmark model in the experiment and achieve state of the art. The model can effectively model threat intelligence text data and acquire knowledge and experience migration by correlating implied features with a priori knowledge to compensate for insufficient sample data and improve the classification performance and recognition ability of threat behavior in text.

Simulation and statistical analysis of ground snow loads based on a multi-layer snow accumulation and melt model

Ground snow loads, the study of which relies on the snowpack data from historical records of meteorological stations, are the basis of calculating the roof snow loads. In the absence of observation data, the snow accumulation and melt model, also called the snowmelt model, can be used to simulate the annual extreme samples of ground snow loads in each locality, and then the ground snow loads under a given recurrence interval can be obtained through probability analysis. In this paper, a multi-layer snowmelt model is used to input meteorological data such as precipitation, air temperature, wind speed, and relative humidity to obtain detailed information on the long-term snowpack, including snow density, snow depth, and snow water equivalent (SWE), which overcomes the problem of insufficient sample data for observed snow loads. Firstly, the precision of multi-layer snowmelt is checked through data of snow depth and SWE. Then, based on the model, meteorological stations in thirty-seven localities of six Chinese regions are selected to simulate the ground snow loads for nearly 60 years. Subsequently, the maximum snow depth and its occurrence time are compared, and the relationship between snow depth, snow density, and SWE in each region is analyzed. Finally, a variety of probabilistic models are used to calculate the ground snow loads at 50-year recurrence intervals. The goodness of fit of each probabilistic model for the snow loads sample is examined, and the calculated values of ground snow loads are compared with recommended values by the Chinese code. The results indicate that the snowmelt model can accurately reproduce the detailed process of ground snow events, which can be used to obtain the samples of ground snow load extreme. Due to gravity compaction, ground snow loads corresponding to the maximum depth in a certain year differ from the maximum ground snow loads in that year, which requires more attention. The goodness of fit is higher and the results are larger than the values of the code when generalized extreme distribution and logarithmic normal distribution are adopted to estimate extreme ground snow loads.

Landslide Displacement Prediction Based on Transfer Learning and Bi-GRU

Predicting slope deformation prediction is crucial for early warning of slope failure, preventing damage to properties, and saving human lives. However, in practice, equipment maintenance causes discontinuity in the displacement data, and the traditional prediction models based on deep networks do not perform well in this case. To solve the problem of prediction accuracy in case of discontinuous and inadequate data, we propose a combined displacement prediction model that integrates the bidirectional gated recurrent unit (Bi-GRU), attention mechanism, and transfer learning. The Bi-GRU is employed to extract the forward and backward characteristics of displacement series, and the attention mechanism is utilized to give different weights to the extracted information so as to highlight the critical information. Transfer learning is used to guarantee prediction accuracy in case of discontinuous and limited data. The model is then employed to predict the slope displacement of the JinYu Cement Plant in China. Finally, the modeling results excellently agree with measured displacement, especially in case of insufficient sample data.

A Novel Sparse Enhancement Neural Network for Rolling Bearing Fault Diagnosis

To ensure the operational reliability of machinery, rolling bearings exposed to complex and poor conditions should be monitored in real-time. Traditional bearing fault diagnosis methods are always dependent on signal analysis and feature extraction, which are complex and time-consuming. Deep learning method exhibits a good ability in extracting the fault feature, while it is limited to noise pollution and insufficient sample data during the training procedure. In this study, a new sparse enhancement neural network based on generalized minimax-concave penalty and convolutional neural network is proposed to capture fault features automatically. To this end, the generalized minimax-concave penalty is first employed to expand the dataset by pollution data denoise and sparse enhancement of the insufficient samples. Second, the amplified dataset is employed to train the fault classification. By employing the datasets of drive end and fan end derived from the Case Western Reserve University (CWRU), a good prediction accuracy can be found in fault diagnosis for rolling bearings.

Impact of Black Summer 2019/20 Wildfires on True Bug Priority Species (Insecta: Hemiptera: Heteroptera) in the Northeast Forests of New South Wales

ABSTRACT This study examined the impact of the Black Summer 2019/20 wildfires in the Northeast Forests of New South Wales for seven priority heteropteran (= true bug) species, across a range of vegetation classes and host plants. Our area of study spanned 104,000 km2, from the Queensland border to the Hunter Valley, and west towards Narrabri. We used data primarily from the Plant Bug Inventory database, inclusive of all phytophagous families of the suborder Heteroptera. We identified 122 survey sites based on their distribution in historical collection data and that of their known host plants. These sites were parsed into unburnt (69 locations) and burnt (53 locations), with the latter further categorised by fire severity. The 122 sites were surveyed in four field trips in the late summer early autumn and spring of 2021. One or more of the priority species were detected at thirty-three sites (22 unburnt and 11 burnt sites). The first three field trips (January–March) resulted in the discovery of only Woodwardiola ‘n.sp_LOMA’. The fourth field trip (November) resulted in the discovery of four additional species (Kirkaldyella rugosa, Setocoris ‘n.sp_BINA’, Epimixia vulturna and Eritingis trivirgata), as well as W. ‘n.sp_LOMA’. The two other priority species (Kirkaldyella schuhi and Myrmecoroides grossi) were not found. IUCN Red list methodology was employed to analyse the conservation status of the seven priority species. All seven priority species were found to be species of Least Concern for EEO and Data Deficient for AOO. The Data Deficient categorisation was decided based on insufficient sampling data plus expert opinion. Further sampling is required for all species but is critical for Kirkaldyella schuhi and Myrmecoroides grossi to determine whether their non-detection in this survey is due to sampling inadequacy or due to environmental factors.

Data Augmentation-aided Convolutional Neural Network for Detection of Abnormalities in Digital Mammography

Background: The use of data augmentation techniques to addressing the challenge of network overfitting and classification error is important in deep learning. Insufficient sample data for training have the tendency to bias the trained model so that it fails to generalize well. Several studies have proposed different augmentation techniques to solve this problem. But there are some peculiarities identified with the nature of datasets when applying augmentation methods. The subtle nature of some abnormalities in digital mammography often makes it difficult to transform such datasets into different form, while preserving the structure of the abnormality. Aim: To address this, this study aims to apply a combination of carefully selected data augmentation operations on digital mammography.