Search results

Genre classification of songs based on lyrics is a challenging task even for humans, however, state-of-the-art natural language processing has recently offered advanced solutions to this task. The purpose of this study is to advance the understanding and application of natural language processing and deep learning in the domain of music genre classification, while also contributing to the broader themes of global knowledge and communication, and sustainable preservation of cultural heritage.

The main contribution of this study is the development and evaluation of various machine and deep learning models for song genre classification. Additionally, we investigated the effect of different word embeddings, including Global Vectors for Word Representation (GloVe) and Word2Vec, on the classification performance. The tested models range from benchmarks such as logistic regression, support vector machine and random forest, to more complex neural network architectures and transformer-based models, such as recurrent neural network, long short-term memory, bidirectional long short-term memory and bidirectional encoder representations from transformers (BERT).

The authors conducted experiments on both English and multilingual data sets for genre classification. The results show that the BERT model achieved the best accuracy on the English data set, whereas cross-lingual language model pretraining based on RoBERTa (XLM-RoBERTa) performed the best on the multilingual data set. This study found that songs in the metal genre were the most accurately labeled, as their text style and topics were the most distinct from other genres. On the contrary, songs from the pop and rock genres were more challenging to differentiate. This study also compared the impact of different word embeddings on the classification task and found that models with GloVe word embeddings outperformed Word2Vec and the learning embedding layer.

This study presents the implementation, testing and comparison of various machine and deep learning models for genre classification. The results demonstrate that transformer models, including BERT, robustly optimized BERT pretraining approach, distilled bidirectional encoder representations from transformers, bidirectional and auto-regressive transformers and XLM-RoBERTa, outperformed other models.

According to the problem of crude oil price forecasting, the purpose of this paper is to propose a multi-step prediction method based on the empirical mode decomposition, long short-term memory network and GM (1,1) model.

First, the empirical mode decomposition method is used to decompose the crude oil price series into several components with different frequencies. Then, each subsequence is classified and synthesized based on the specific periodicity and other properties to obtain several components with different significant characteristics. Finally, all components are substituted into a suitable prediction model for fitting. LSTM models with different parameters are constructed for predicting specific components, which approximately and respectively represent short-term market disturbance and long-term influences. Rolling GM (1,1) model is constructed to simulate a series representing the development trend of oil price. Eventually, all results obtained from forecasting models are summarized to evaluate the performance of the model.

The model is respectively applied to simulate daily, weekly and monthly WTI crude oil price sequences. The results show that the model has high accuracy on the prediction, especially in terms of series representing long-term influences with lower frequency. GM (1,1) model has excellent performance on fitting the trend of crude oil price.

This paper combines GM (1,1) model with LSTM network to forecast WTI crude oil price series. According to the different characteristics of different sequences, suitable forecasting models are constructed to simulate the components.

This paper aims to develop predictive methods, based on recurrent neural networks, useful to support facility managers in building maintenance tasks, by collecting information coming from a computerized maintenance management system (CMMS).

This study applies data-driven and text-mining approaches to a CMMS data set comprising more than 14,500 end-users’ requests for corrective maintenance actions, collected over 14 months. Unidirectional long short-term memory (LSTM) and bidirectional LSTM (Bi-LSTM) recurrent neural networks are trained to predict the priority of each maintenance request and the related technical staff assignment. The data set is also used to depict an overview of corrective maintenance needs and related performances and to verify the most relevant elements in the building and how the current facility management (FM) relates to the requests.

The study shows that LSTM and Bi-LSTM recurrent neural networks can properly recognize the words contained in the requests, thus correctly and automatically assigning the priority and predicting the technical staff to assign for each end-user’s maintenance request. The obtained global accuracy is very high, reaching 93.3% for priority identification and 96.7% for technical staff assignment. Results also show the main critical building elements for maintenance requests and the related intervention timings.

This work shows that LSTM and Bi-LSTM recurrent neural networks can automate the assignment process of end-users’ maintenance requests if trained with historical CMMS data. Results are promising; however, the trained LSTM and Bi-LSTM RNN can be applied only to different hospitals adopting similar categorization.

The data-driven and text-mining approaches can be integrated into the CMMS to support corrective maintenance management by facilities management contractors, i.e. to properly and timely identify the actions to be carried out and the technical staff to assign.

The improvement of the maintenance of the health-care system is a key component of improving health service delivery. This work shows how to reduce health-care service interruptions due to maintenance needs through machine learning methods.

This study develops original methods and tools easily integrable into IT workflow systems (i.e. CMMS) in the FM field.

The purpose of this paper is to propose a distributed deep learning architecture for smart cities in big data systems.

We have proposed an architectural multilayer to describe the distributed deep learning for smart cities in big data systems. The components of our system are Smart city layer, big data layer, and deep learning layer. The Smart city layer responsible for the question of Smart city components, its Internet of things, sensors and effectors, and its integration in the system, big data layer concerns data characteristics 10, and its distribution over the system. The deep learning layer is the model of our system. It is responsible for data analysis.

We apply our proposed architecture in a Smart environment and Smart energy. 10; In a Smart environment, we study the Toluene forecasting in Madrid Smart city. For Smart energy, we study wind energy foresting in Australia. Our proposed architecture can reduce the time of execution and improve the deep learning model, such as Long Term Short Memory10;.

This research needs the application of other deep learning models, such as convolution neuronal network and autoencoder.

Findings of the research will be helpful in Smart city architecture. It can provide a clear view into a Smart city, data storage, and data analysis. The 10; Toluene forecasting in a Smart environment can help the decision-maker to ensure environmental safety. The Smart energy of our proposed model can give a clear prediction of power generation.

The findings of this study are expected to contribute valuable information to decision-makers for a better understanding of the key to Smart city architecture. Its relation with data storage, processing, and data analysis.

Decision-making in human beings is affected by emotions and sentiments. The affective computing takes this into account, intending to tailor decision support to the emotional states of people. However, the representation and classification of emotions is a very challenging task. The study used customized methods of deep learning models to aid in the accurate classification of emotions and sentiments.

The present study presents affective computing model using both text and image data. The text-based affective computing was conducted on four standard datasets using three deep learning customized models, namely LSTM, GRU and CNN. The study used four variants of deep learning including the LSTM model, LSTM model with GloVe embeddings, Bi-directional LSTM model and LSTM model with attention layer.

The result suggests that the proposed method outperforms the earlier methods. For image-based affective computing, the data was extracted from Instagram, and Facial emotion recognition was carried out using three deep learning models, namely CNN, transfer learning with VGG-19 model and transfer learning with ResNet-18 model. The results suggest that the proposed methods for both text and image can be used for affective computing and aid in decision-making.

The study used deep learning for affective computing. Earlier studies have used machine learning algorithms for affective computing. However, the present study uses deep learning for affective computing.

In the era of artificial intelligence (AI), algorithms have gained unprecedented importance. Scientific studies have shown that algorithms are frequently mentioned in papers, making mention frequency a classical indicator of their popularity and influence. However, contemporary methods for evaluating influence tend to focus solely on individual algorithms, disregarding the collective impact resulting from the interconnectedness of these algorithms, which can provide a new way to reveal their roles and importance within algorithm clusters. This paper aims to build the co-occurrence network of algorithms in the natural language processing field based on the full-text content of academic papers and analyze the academic influence of algorithms in the group based on the features of the network.

We use deep learning models to extract algorithm entities from articles and construct the whole, cumulative and annual co-occurrence networks. We first analyze the characteristics of algorithm networks and then use various centrality metrics to obtain the score and ranking of group influence for each algorithm in the whole domain and each year. Finally, we analyze the influence evolution of different representative algorithms.

The results indicate that algorithm networks also have the characteristics of complex networks, with tight connections between nodes developing over approximately four decades. For different algorithms, algorithms that are classic, high-performing and appear at the junctions of different eras can possess high popularity, control, central position and balanced influence in the network. As an algorithm gradually diminishes its sway within the group, it typically loses its core position first, followed by a dwindling association with other algorithms.

To the best of the authors’ knowledge, this paper is the first large-scale analysis of algorithm networks. The extensive temporal coverage, spanning over four decades of academic publications, ensures the depth and integrity of the network. Our results serve as a cornerstone for constructing multifaceted networks interlinking algorithms, scholars and tasks, facilitating future exploration of their scientific roles and semantic relations.

Building energy management systems use important information from indoor room temperature (IRT) forecasting to predict daily loads within smart buildings. IRT forecasting is a complex and challenging task, especially when energy demands are exponentially rising. The purpose of this paper is to review the relevant literature on indoor temperature forecasting in the past two decades and draw inferences on important methodologies with influencing variables and offer future directions.

The motivation for this work is based on the research work done in the field of intelligent buildings and energy related sector. The focus of this study is based on past literature on forecasting models and methodologies related to IRT forecasting for building energy management, with an emphasis on data-driven models (statistical and machine learning models). The methodology adopted here includes review of several journals, conference papers, reference books and PhD theses. Selected forecasting methodologies have been reviewed for indoor temperature forecasting contributing to building energy consumption. The models reviewed here have been earmarked for their benefits, limitations, location of study, accuracy along with the identification of influencing variables.

The findings are based on 62 studies where certain accuracy metrics and influencing explanatory variables have been reviewed. Linear models have been found to show explanatory relationships between the variables. Nonlinear models are found to have better accuracy than linear models. Moreover, IRT profiles can be modeled with enhanced accuracy and generalizability through hybrid models. Although deep learning models are found to have better performance for this study.

This is accuracy-based study of data-driven models. Their run-time performance and cost implications review and review of physical, thermal and simulation models is future scope.

Despite the earlier work conducted in this field, there is a lack of organized and comprehensive evaluation of peer reviewed forecasting methodologies. Indoor temperature depends on various influencing explanatory variables which poses a research challenge for researchers to develop suitable predictive model. This paper presents a critical review of selected forecasting methodologies and provides a list of important methodologies along with influencing variables, which can help future researchers in the field of building energy management sector. The forecasting methods presented here can help to determine appropriate heating, ventilation and air-conditioning systems for buildings.

Tourism spot or point-of-interest (POI) recommendation has become a common service in people's daily life. The purpose of this paper is to model users' check-in history in order to predict a set of locations that a user may soon visit.

The authors proposed a novel learning-based method, the pattern-based dual learning POI recommendation system as a solution to consider users' interests and the uniformity of popular POI patterns when making recommendations. Differing from traditional long short-term memory (LSTM), a new users’ regularity–POIs’ popularity patterns long short-term memory (UP-LSTM) model was developed to concurrently combine the behaviors of a specific user and common users.

The authors introduced the concept of dual learning for POI recommendation. Several performance evaluations were conducted on real-life mobility data sets to demonstrate the effectiveness and practicability of POI recommendations. The metrics such as hit rate, precision, recall and F-measure were used to measure the capability of ranking and precise prediction of the proposed model over all baselines. The experimental results indicated that the proposed UP-LSTM model consistently outperformed the state-of-the-art models in all metrics by a large margin.

This study contributes to the existing literature by incorporating a novel pattern–based technique to analyze how the popularity of POIs affects the next move of a particular user. Also, the authors have proposed an effective fusing scheme to boost the prediction performance in the proposed UP-LSTM model. The experimental results and discussions indicate that the combination of the user's regularity and the POIs’ popularity patterns in PDLRec could significantly enhance the performance of POI recommendation.

The aim of this study is to evaluate of the predictive ability of goodwill and other intangible assets on forecasting corporate profitability. Subsequently, this study compares the efficiency of deep learning model to that of other machine learning models such as random forest (RF) and support vector machine (SVM) as well as traditional statistical methods such as the linear regression model.

Studies confirm that goodwill and intangibles are valuable assets that give companies a competitive advantage to increase profitability and shareholders’ returns. Thus, by using as sample Greek-listed financial data, this study investigates whether or not the inclusion of goodwill and intangible assets as input variables in this modified deep learning models contribute to the corporate profitability prediction accuracy. Subsequently, this study compares the modified long-short-term model with other machine learning models such as SVMs and RF as well as the traditional panel regression model.

The findings of this paper confirm that goodwill and intangible assets clearly improve the performance of a deep learning corporate profitability prediction model. Furthermore, this study provides evidence that the modified long short-term memory model outperforms other machine learning models such as SVMs and RF , as well as traditional statistical panel regression model, in predicting corporate profitability.

Limitation of this study includes the relatively small amount of data available. Furthermore, the aim is to challenge the authors’ modified long short-term memory by using listed corporate data of Greece, a code-law country that suffered severely during the recent fiscal crisis. However, this study proposes that future research may apply deep learning corporate profitability models on a bigger pool of data such as STOXX Europe 600 companies.

Subsequently, the authors believe that their paper is of interest to different professional groups, such as financial analysts and banks, which the authors’ paper can support in their corporate profitability evaluation procedure. Furthermore, as well as shareholders are concerned, this paper could be of benefit in forecasting management’s potential to create future returns. Finally, management may incorporate this model in the evaluation process of potential acquisitions of other companies.

The contributions of this work can be summarized in the following aspects. This study provides evidence that by including goodwill and other intangible assets in the authors’ input portfolio, prediction errors represented by root mean squared error are reduced. A modified long short-term memory model is proposed to predict the numerical value of the profitability (or the profitability ratio) in contrast to other studies which deal with trend predictions, i.e. the binomial output result of positive or negative earnings. Finally, posing an extra challenge to the authors’ deep learning model, the authors’ used financial statements according to International Financial Reporting Standard data of listed companies in Greece, a code-law country that suffered during the recent fiscal debt crisis, heavily influenced by tax legislation and characterized by its lower investors’ protection compared to common-law countries.

An intelligent personal assistant for personal computers (PCs) is a vital application for the current generation. The current computer personal assistant services checking frameworks are not proficient at removing significant data from PCs and long-range informal communication information.

The proposed verbalizers use long short-term memory to classify the user task and give proper guidelines to the users. The outcomes show that the proposed method determinedly handles heterogeneous information and improves precision. The main advantage of long short-term memory is that handle the long-term dependencies in the input data.

The proposed model gives the 22% mean absolute error. The proposed method reduces mean square error than support vector machine (SVM), convolutional neural network (CNN), multilayer perceptron (MLP) and K-nearest neighbors (KNN).

This paper fulfills the necessity of intelligent personal assistant for PCs using verbalizer.