Classification and pattern extraction of incidents: a deep learning-based approach

Abstract

Classifying or predicting occupational incidents using both structured and unstructured (text) data are an unexplored area of research. Unstructured texts, i.e., incident narratives are often unutilized or underutilized. Besides the explicit information, there exist a large amount of hidden information present in a dataset, which cannot be explored by the traditional machine learning (ML) algorithms. There is a scarcity of studies that reveal the use of deep neural networks (DNNs) in the domain of incident prediction, and its parameter optimization for achieving better prediction power. To address these issues, initially, key terms are extracted from the unstructured texts using LDA-based topic modeling. Then, these key terms are added with the predictor categories to form the feature vector, which is further processed for noise reduction and fed to the adaptive moment estimation (ADAM)-based DNN (i.e., ADNN) for classification, as ADAM is superior to GD, SGD, and RMSProp. To evaluate the effectiveness of our proposed method, a comparative study has been conducted using some state-of-the-arts on five benchmark datasets. Moreover, a case study of an integrated steel plant in India has been demonstrated for the validation of the proposed model. Experimental results reveal that ADNN produces superior performance than others in terms of accuracy. Therefore, the present study offers a robust methodological guide that enables us to handle the issues of unstructured data and hidden information for developing a predictive model.

Phenotypic Analysis of Diseased Plant Leaves Using Supervised and Weakly Supervised Deep Learning

“…The value of batch size was set according to the memory size of the GPU. The Adam [ 21 ] method was selected as the optimizer for training, targeting to minimize the fused loss function (binary cross-entropy loss and dice loss [ 22 ]) between the prediction and ground truth, which could be expressed as: where L Dice denotes the dice loss, L BCE is the loss of binary cross-entropy with logits, y p and y are the predicted value and ground truth, ε is a small value to ensure that the denominator is a non-zero value, σ is the sigmoid function, λ is the weight of L BCE , and L is the final loss function of the segmentation models. λ was set as 0.5 in this study.…”

Contextualizing injury severity from occupational accident reports using an optimized deep learning prediction model

“…Moreover, the inclusion of “trapped” suggests situations where body parts are confined, possibly in machinery or equipment, which can lead to severe injuries that may necessitate amputation. Based on the interpretability analysis, it is evident that the occupational injury narratives contained keywords that delineated the accident’s type or causes ( Sarkar et al, 2022 ), as well as the affected body parts, along with the nature of the injury or outcomes ( Davoudi Kakhki, Freeman & Mosher, 2019 ; Kang, Koo & Ryu, 2022 ; Yedla, Kakhki & Jannesari, 2020 ). These findings align with those of similar studies in the field, providing comparable insights into the predictors and consequences of occupational injuries.…”

“…Expanding the analysis to incorporate additional modalities such as occupational injury images or audio data from accident investigations could contribute to a more comprehensive knowledge of the severity of occupational injuries. Multimodal approaches have the potential to capture richer contextual information and improve the prediction performance ( Sarkar et al, 2022 ).…”

Transformer-based Deep Learning for COVID-19 Prediction Based on Climate Variables in Indonesia

“…Adaptive Moment Estimation (ADAM) was selected as the optimizer for this training. ADAM is the combined result of momentum optimization [33] and RMSProp [34]. ADAM's advantage is that convergence and more efficient computing accelerates the training process of the deep learning model [35].…”