Sistema Regional de Información
en línea para Revistas Científicas de América Latina,
el Caribe, España y Portugal

This study examines the prevalence of big bath accounting in an emerging market context, focusing on newly appointed CEOs of Brazilian firms. We find evidence of big bath accounting in Brazil and extends the big bath and CEO turnover literature by documenting a limit to downward earnings manipulation and by using Brazil as a source of study. Our results suggest that incumbent CEOs have incentives to manage profits downward only when the company was previously profitable, rather than risking worsening an already bad situation, in the case of unprofitable firms. We also find that they are more likely to use accrual earnings management, rather than reducing production or increasing discretionary expenditures, as tool to decrease earnings level. Overall, our study fills a gap in the literature and supports future research in understanding the big bath CEO turnover scenario in Brazilian firms.  

Lamentamos profundamente el repentino fallecimiento del distinguido profesor Jorge Heredia Pérez. El Dr. Heredia, académico del Departamento de Gestión y Negocios de la Universidad Alberto Hurtado y miembro del equipo editorial del Journal of Technology Management and Innovation, nos dejó una profunda huella, no solo por su aporte a la revista, sino también por su pasión, compromiso y contribuciones en temas de innovación, management y estrategia.

Nowadays, Machine learning solutions are increasing their presence in the industry, and the benefits associated with the use of this technology are reflected in a reduction of time, costs, and a clear economic benefit, also most of these solutions use supervised learning, where a dataset with labeled real examples is needed.  However, the many challenges associated with the implementation of a supervised machine learning solution are sometimes difficult to overpass, so one of the most important challenges is the creation of a big labeled dataset to feed the algorithms, since most of the time there is no any dataset available for the proposed solution, and most companies cannot afford to have a hand-created dataset with tens of thousands of records, also because they even cannot be sure that the model will work. And the problem is that current machine learning techniques fail to improve the problem understanding on training with very small datasets, so here it is shown how by using a novel XAI method used to explain the decisions of a machine learning model in a computer vision problem, we can augment the labeled dataset focused on the important regions for the image classification, and increase the model performance, not just for training but for validation and testing, also this method turns out to be superior to the most used data augmentation methods if we further reduce the amount of information.

In the biological area, the short reproductive cycle in rodents is useful because it allows analyzed electrophysiological properties, behaviors, or drug effects through the changes observed during this period. This cycle is composed of 4 stages in which the classification is determined by vaginal cytology. Although automatic approaches have been used for the classification of these stages, they are computationally expensive and require a great number of images for adequate performance. In this paper, we test different models of dendritic neural networks (DNN) trained by stochastic gradient descent to classify a short number of images and four classical contrast enhancement methods. We extract texture features and use standard and DNN classifiers to recognize the images. From the experiments, it seems that DNNs have a more stable behavior concerning the standard classifiers according to the standard deviation presented, being this a desirable property for a model. We consider that DNN could be an adequate alternative for the classification of estrous cycle images.

This paper reviews the implementation of three pre-trained models based on BERT (“bert-base-multilingual-cased”, “IIC/beto-base-spanish-sqac” and “MarcBrun/ixambert-finetuned-squad-eu-en”) to solve tasks 1.1 and 1.2 of “Workshop on Semantic Analysis at SEPLN 2020” (TASS 2020), these tasks consist of the polarity analysis of tweets in Spanish from different Spanish-speaking countries. The proposed models are evaluated individually by pre-processing and replacing synonyms. This research is carried out to find the points to improve in the polarity analysis of tweets (tweets), mainly in how the pre-trained models interpret words that are not in their vocabulary due to variations in the language, regional expressions, misspellings, and use of emojis.

The microclimate inside a greenhouse forecast has been a case of study in recent years; an adequate forecast of variables such as internal temperature helps farmers prevent losses in the harvest. In this investigation, the forecast of the greenhouse internal temperature is implemented through Recurrent Neural Networks (RNN) topology with Long-Short Term Memory (LSTM) algorithm. The analysis is performed with the many to one configuration for a sequence of three input elements and one output element for each of the year's four seasons. The metrics used for the analysis and validation of the data were the RMSE, MAE, R2, and Ceff. These metrics provide the level of efficiency and goodness of the RNN-LSTM showing how the variables considered provide significance to the forecast of one hour into the future of the internal temperature. It is shown that the LSTM algorithm within the RNN is an effective tool for a good internal temperature forecast in time series for each season, significantly helping the forecast of climatic variables inside a greenhouse.

There exist multiple online collections and data bases of scientific articles publicly available, to take full advantage of these resources, it is necessary to process, arrange and correlate texts with respect to a classification or ontology. To achieve an efficient organization and a more relevant correlation between texts, it is necessary to use a similarity measure for short texts. However, determining the best method to calculate the similarity between texts is an arduous task, since there are many similarity measures reported in literature. Additionally, the collection of texts to which the similarity measures are applied should be considered; while some measures are useful for some types of information sources, they fail when the collection of data changes. Therefore, it is necessary to count with a method to evaluate the performance of similarity measures from a statistical perspective and in terms of the accuracy achieved by each measure.

Parking lot systems based on computer vision have been used frequently in recent years to improve vehicle traffic on urban areas. These systems provide information about the availability of a parking place, furthermore these systems contribute for a better organization and to reduce the time to looking for a free space. One of the main challenges for these systems is the occlusion effects that occur frequently due to the location of the cameras in the parking lots. In this paper, we present a method for available parking place detection, where we contribute with a mechanism to reduce occlusion effects that consists of feature extraction using ICM based on input image separation. The results shows that our contribution significantly reduces the noise generated by occlusion effects. The proposed method was evaluated by our dataset and an external dataset, where the experiments results achieving up 0.98 accuracy.

We show how properties of the sequence βi,j, which represents the product between two Fibonacci's numbers Fi × Fj, can be used for the computation of the Merrifield-Simmons index on bipolygonal graphs and trees. We show that the extreme values of the Merrifield-Simmons index on bipolygonal graphs are found in two consecutive columns of the table βi,j k=i+j=1,...,n. The minimum value in β3,k-3 and the maximum value in β4,k-4.  On the other hand we show that i(Tn ∪ {{vp, v}}) is minimum when v is a new leaf node, and its father vp was also a leaf node in Tn. Our methods does not require the explicit computation of the number of independent sets of the involved graphs. Instead, it is based on applying the edge and vertex division rules to decompose the initial graph.

One of the most important diseases worldwide in public health is Diabetes Mellitus (DM) since this is one of the most severe and frequent non-communicable diseases with various chronic complications. In this paper, we propose a procedure to detect the most common risk factors in patients suffering from the disease known as diabetes mellitus, through principal components analysis (PCA) and non-negative matrix factorization (NMF). We then check the results using these factors like features, through the machine learning algorithms, improving the classification results. According to the experimental results, accuracy of more than 80% was obtained.