Business transaction recommendation for discovering potential business partners using deep learning

Prachyachuwong K., Vateekul P.

A stock trend prediction has been in the spotlight from the past to the present. Fortunately, there is an enormous amount of information available nowadays. There were prior attempts that have tried to forecast the trend using textual information; however, it can be further improved since they relied on fixed word embedding, and it depends on the sentiment of the whole market. In this paper, we propose a deep learning model to predict the Thailand Futures Exchange (TFEX) with the ability to analyze both numerical and textual information. We have used Thai economic news headlines from various online sources. To obtain better news sentiment, we have divided the headlines into industry-specific indexes (also called “sectors”) to reflect the movement of securities of the same fundamental. The proposed method consists of Long Short-Term Memory Network (LSTM) and Bidirectional Encoder Representations from Transformers (BERT) architectures to predict daily stock market activity. We have evaluated model performance by considering predictive accuracy and the returns obtained from the simulation of buying and selling. The experimental results demonstrate that enhancing both numerical and textual information of each sector can improve prediction performance and outperform all baselines.

Yang Z., Baraldi P., Zio E.

• We use numerical data, images and texts for fault prognostics. • The method is able to extract prognostic features from multimodal data. • The method is applied to multimodal data obtained from steam generators. • The use of multimodal data allows obtaining more accurate predictions. Non-numerical data, such as images and inspection records, contain information about industrial system degradation, but they are rarely used for failure prognostic tasks given the difficulty of automatic analysis. In this work, we present a novel method for prognostics using multimodal data, i.e. both numerical and non-numerical data. The proposed method is based on the development of a multi-branch Deep Neural Network (DNN), each branch of which is a neural network designed for processing a certain type of data. The method is applied to a case study properly designed to reproduce the problem of prognostics using multimodal data by referring to the operation of steam generators. The results show that it is able to accurately predict future degradation level using multimodal data, outperforming other methods using fewer sources of information.

Madeh Piryonesi S., El-Diraby T.E.

AbstractLimited research has been conducted on the application of data analytics to the prediction of the Pavement Condition Index (PCI) of asphalt roads. More importantly, studies comparing the pr...

Bach T., Harvie C., Le T.

This study makes use of rich firm-level and linked firm-employee datasets that span the 2009–2015 period in Vietnam to examine how SMEs' credit constraints affect their strategic employment decisions and employees' labour outcomes. Our results show that constrained SMEs enlarge total employment by employing relatively more temporary workers and paying their employees relatively lower wages than unconstrained borrowing firms. Meanwhile, discouraged firms, mostly informal businesses, do not behave differently from unconstrained counterparts. In order to maintain a stable employment portfolio, discouraged firms are relatively more willing to reward their employees with an overtime payment.

Kim Y.J., Kim S.

Xiong X., Xiong F., Zhao J., Qiao S., Li Y., Zhao Y.

A large volume of data flowing throughout location-based social networks (LBSN) gives support to the recommendation of points-of-interest (POI). One of the major challenges that significantly affects the precision of recommendation is to find dynamic spatio-temporal patterns of visiting behaviors, which can hardly be figured out because of the multiple side factors. To confront this difficulty, we jointly study the effects of users’ social relationships, textual reviews, and POIs’ geographical proximity in order to excavate complex spatio-temporal patterns of visiting behaviors when the data quality is unreliable for location recommendation in spatio-temporal social networks. We craft a novel framework that recommends any user the POIs with effectiveness. The framework contains two significant techniques: (i) a network embedding method is adopted to learn the vectors of users and POIs in an embedding space of low dimension; (ii) a dynamic factor graph model is proposed to model various factors such as the correlation of vectors in the previous phase. A collection of experiments was carried out on two real large-scale datasets, and the experimental outcomes demonstrate the supremacy of the proposed method over the most advanced baseline algorithms owing to its highly effective and efficient performance of POI recommendation.

Wang R., Jiang Y., Lou J.

Recently, deep learning techniques have been widely used in recommendation tasks and have attained record performance. However, the input quality of the deep learning model has a great influence on the recommendation performance. In this work, an efficient and effective input optimization method is proposed. Specifically, we propose an integrated recommendation framework based on two-stage deep learning. In the first stage, with user and item features as the original input, a low-cost marginalized stacked denoising auto-encoder (mSDA) model is used to learn the latent factors of users and items. In the second stage, the resulting latent factors are combined and used as input vector to the DNN model for fast and accurate prediction. Using the latent factor vector as the input to the deep learning-based recommendation model not only captures the high-order feature interaction, but also reduces the burden of the hidden layer, and also avoids the model training falling into local optimum. Extensive experiments with real-world datasets show that the proposed model shows much better performance than the state-of-the-art recommendation methods in terms of prediction accuracy, parameter space and training speed.

Lee D., Kim K.

Recently, the prediction of photovoltaic (PV) power has become of paramount importance to improve the expected revenue of PV operators and the effective operations of PV facility systems. Additionally, the precise PV power output prediction in an hourly manner enables more sophisticated strategies for PV operators and markets as the electricity price in a renewable energy market is continuously changing. However, the hourly prediction of PV power outputs is considered as a challenging problem due to the dynamic natures of meteorological information not only in a day but also across days. Therefore, in this paper, we suggest three PV power output prediction methods such as artificial neural network (ANN)-, deep neural network (DNN)-, and long and short term memory (LSTM)-based models that are capable to understand the hidden relationships between meteorological information and actual PV power outputs. In particular, the proposed LSTM based model is designed to capture both hourly patterns in a day and seasonal patterns across days. We conducted the experiments by using a real-world dataset. The experimental results show that the proposed ANN based model fails to yield satisfactory results, and the proposed LSTM based model successfully better performs more than 50% compared to the conventional statistical models in terms of mean absolute error.

Messina P., Dominguez V., Parra D., Trattner C., Soto A.

Recommender Systems help us deal with information overload by suggesting relevant items based on our personal preferences. Although there is a large body of research in areas such as movies or music, artwork recommendation has received comparatively little attention, despite the continuous growth of the artwork market. Most previous research has relied on ratings and metadata, and a few recent works have exploited visual features extracted with deep neural networks (DNN) to recommend digital art. In this work, we contribute to the area of content-based artwork recommendation of physical paintings by studying the impact of the aforementioned features (artwork metadata, neural visual features), as well as manually-engineered visual features, such as naturalness, brightness and contrast. We implement and evaluate our method using transactional data from UGallery.com, an online artwork store. Our results show that artwork recommendations based on a hybrid combination of artist preference, curated attributes, deep neural visual features and manually-engineered visual features produce the best performance. Moreover, we discuss the trade-off between automatically obtained DNN features and manually-engineered visual features for the purpose of explainability, as well as the impact of user profile size on predictions. Our research informs the development of next-generation content-based artwork recommenders which rely on different types of data, from text to multimedia.

Choi Y., Kwak D., Jung M., Lee D.H., Park S., Kim K.

Ding R., Chen Z.

How to exploit various features of users and points of interest (POIs) for accurate POI recommendation is important in location-based social networks (LBSNs). In this paper, a novel POI recommendat...

Alberti F.G., Varon Garrido M.A.

Purpose This paper aims to discuss hybrid organizations whose business models blur the boundary between for-profit and nonprofit worlds. With the aim of understanding how hybrid organizations have developed commercially viable business models to create positive social and environmental change, the authors contend that hybrids are altering long-held business norms and conceptions of the role of the corporation in society. Building on an analysis of the most updated literature on hybrid organizations and with the use of case study approach, the purpose of this paper is to derive managerial lessons that traditional businesses may apply to innovate their business models. Design/methodology/approach This paper has a practical focus to help organizations to develop successful business strategies and design innovative business models. It applies emerging thinking on hybrid business models to provide new insights and ideas on the use of business models as tools for innovating and delivering value. To comply with this, first, the authors discuss the distinctive characteristics of hybrids and the hybrid business model through a concise but comprehensive review of all the literature on hybrid organization, which is still very recent. Second, we relied on a short case study that introduces information technology and digital innovation as the premises of the emergence of a new hybrid business model that adds additional elements to traditional business managers on how to learn from hybrid organizations’ avenues to innovate their business models. Findings In this paper, the authors aimed to shed light on the management of any organization or initiative that aims to embrace multiple and competing yet potentially synergistic goals, as is increasingly the case in modern corporations. Spotting hidden complementarities of antagonistic assets can be arduous, time-consuming, costly and risky, but businesses driven by innovation may want to keep a close eye on the expanding hybrid sector as a source of future entrepreneurial opportunities. To this regard, hybrid social ventures have the potential to shed light on ways to innovate traditional business models. The essence of studying hybrids is that firms may learn how to innovate their business models in ways that go beyond current conceptualizations, making their mission profitable, rather than making profit their only mission! The research design (literature analysis and case study) allowed the authors to disentangle different innovative business models that hybrids suggest highlight strengths and weaknesses of such business models, understand strategies and capabilities associated with hybrids and transpose all these lessons learned to traditional business managers who constantly struggle for innovation. Research limitations/implications The main implication is that hybrid organizations may serve as incubators for new practices that can gain scale and impact by infusion into existing corporations. The authors can assist to a process of “hybridization” of incumbent firms, pushing the boundaries of corporate sustainability efforts toward strategies in which profit and social purpose share more equal footing. Practical implications Firms interested in benefiting from antagonistic assets that can have a dramatic impact on their business model innovation may want to consider some lessons: firms can attempt to build antagonistic assets into their mission, asking themselves what activities they can undertake with the potential to create (or erode) social, environmental and economic value and how these activities might be mediated by the context/environment in which they operate; they can partner with hybrids to benefit from them and absorb competencies from them, so to increase their likelihood to generate value-creating activities and to impact on wider range of stakeholders, including funders, partners, beneficiaries and communities; they can mimic hybrids on how to innovate their business model through the use of the “deliberate resource misfit” dynamic capability, mitigating negative impacts and trade-offs and maximizing positive value spillovers, both for the firms themselves and for the community. Social implications Sharing know-how with hybrids opens up to ways to innovate business models, and hybrids are much more open to sharing lessons and encouraging others to copy their approaches in a genuine open innovation approach. Originality/value The main lesson businesses can take away from studying hybrids is that antagonistic assets – and not only profitable complementary ones, as the resource-based view would suggest – do not have to be a burden on profits. Hybrids ground their strategy first and foremost on their beneficiaries, thus dealing with a bundle of antagonistic assets. The primary objective of hybrids is thus to find imaginative ways of generating profits from their given resources rather than acquiring the resources that generate the highest profit. Profit is the ultimate goal of traditional businesses’ mission, but by making profit their only mission, firms risk missing out on the hidden opportunities latent in antagonistic assets. Learning from hybrids about how to align profits and societal impact may be a driver of long-term competitive advantage.

Yujun Wen, Hui Yuan, Pengzhou Zhang

In this paper, we do a research on the keyword extraction method of news articles. We build a candidate keywords graph model based on the basic idea of TextRank, use Word2Vec to calculate the similarity between words as transition probability of nodes' weight, calculate the word score by iterative method and pick the top N of the candidate keywords as the final results. Experimental results show that the weighted TextRank algorithm with correlation of words can improve performance of keyword extraction generally.

Wang S., Liu W., Wu J., Cao L., Meng Q., Kennedy P.J.

Deep learning has become increasingly popular in both academic and industrial areas in the past years. Various domains including pattern recognition, computer vision, and natural language processing have witnessed the great power of deep networks. However, current studies on deep learning mainly focus on data sets with balanced class labels, while its performance on imbalanced data is not well examined. Imbalanced data sets exist widely in real world and they have been providing great challenges for classification tasks. In this paper, we focus on the problem of classification using deep network on imbalanced data sets. Specifically, a novel loss function called mean false error together with its improved version mean squared false error are proposed for the training of deep networks on imbalanced data sets. The proposed method can effectively capture classification errors from both majority class and minority class equally. Experiments and comparisons demonstrate the superiority of the proposed approach compared with conventional methods in classifying imbalanced data sets on deep neural networks.

Ma L., Zhang Y.

Big data is a broad data set that has been used in many fields. To process huge data set is a time consuming work, not only due to its big volume of data size, but also because data type and structure can be different and complex. Currently, many data mining and machine learning technique are being applied to deal with big data problem; some of them can construct a good learning algorithm in terms of lots of training example. However, considering the data dimension, it will be more efficient if learning algorithm is capable of selecting useful features or decreasing the feature dimension. Word2Vec, proposed and supported by Google, is not an individual algorithm, but it consists of two learning models, Continuous Bag of Words (CBOW) and Skip-gram. By feeding text data into one of learning models, Word2Vec outputs word vectors that can be represented as a large piece of text or even the entire article. In our work, we first training the data via Word2Vec model and evaluated the word similarity. In addition, we clustering the similar words together and use the generated clusters to fit into a new data dimension so that the data dimension is decreased.

Karbevska L., Hidalgo C.A.

Abstract Value chain data is crucial for navigating economic disruptions. Yet, despite its importance, we lack publicly available product-level value chain datasets, since resources such as the “World Input-Output Database”, “Inter-Country Input-Output Tables”, “EXIOBASE”, and “EORA”, lack information about products (e.g. Radio Receivers, Telephones, Electrical Capacitors, LCDs, etc.) and instead rely on aggregate industrial sectors (e.g. Electrical Equipment, Telecommunications). Here, we introduce a method that leverages ideas from machine learning and trade theory to infer product-level value chain relationships from fine-grained international trade data. We apply our method to data summarizing the exports and imports of 1200+ products and 250+ world regions (e.g. states in the U.S., prefectures in Japan, etc.) to infer value chain information implicit in their trade patterns. In short, we leverage the idea that due to global value chains, regions specialized in the export of a product will tend to specialize in the import of its inputs. We use this idea to develop a novel proportional allocation model to estimate product-level trade flows between regions and countries. This contributes a method to approximate value chain data at the product level that should be of interest to people working in logistics, trade, and sustainable development.

Song S., Kim J., Kim K., Kim J., Lee D.

This study investigates the impact of network capabilities and deep learning techniques on companies' financial performance within buyer-supplier networks. It broadens the scope by incorporating network measures such as closeness and network constraint, whereas previous studies have primarily focused on suitable buyer-supplier relationships. The analysis evaluates the effects of these network measures on companies' financial performance metrics, including asset growth rate, return on assets, and more. In addition, this study explores the impact of extended networks on company performance using deep learning techniques. The results show that companies' network capabilities are positively associated with their financial performance, highlighting the critical role of network positions in driving success. Furthermore, the findings suggest that extending the network through deep learning offers significant benefits for companies.

Mungo L., Brintrup A., Garlaschelli D., Lafond F.

Abstract Network reconstruction is a well-developed sub-field of network science, but it has only recently been applied to production networks, where nodes are firms and edges represent customer-supplier relationships. We review the literature that has flourished to infer the topology of these networks by partial, aggregate, or indirect observation of the data. We discuss why this is an important endeavour, what needs to be reconstructed, what makes it different from other network reconstruction problems, and how different researchers have approached the problem. We conclude with a research agenda.

Ambika N.

Business transactions are a fundamental part of any business or organization, and they play a crucial role in tracking and managing financial activities, maintaining transparency, and ensuring compliance with legal and regulatory requirements. The previous work suggests four stages. The relevant data is collected in this stage. This procedure enhances integrity in the system. It aims to predict the model's accuracy. The collected details are considered as training information. The pre-processing phase does the feature extraction, based on what is expected as the outcome. The feature set is created considering the problem to be addressed. The errors like inconsistencies, redundancy, and missing data are removed. Model building is constructed using machine learning techniques. Model training and testing is divided into two sets. The training dataset is constructed by analyzing the preliminary data input. The test dataset is fed into the system, and the training dataset is compared with the test data to make predictions. The suggestion uses a backpropagation algorithm to make the prediction.

Kumar A., Singh D., Shankar Yadav R.

Class overlap in imbalanced datasets is the most common challenging situation for researchers in the fields of deep learning (DL) machine learning (ML), and big data (BD) based applications. Class overlap and imbalance data intrinsic characteristics negatively affect the performance of classification models. The data level, algorithm level, ensemble, and hybrid methods are the most commonly used solutions to reduce the biasing of the standard classification model towards the majority class. The data level methods change the distribution of class instances thus, increasing the information loss and overfitting. The algorithm-level methods attempt to modify its structure which gives more weight to the misclassified minority class instances in the learning phases. However, the changes in the algorithm are less compatible for the users. To overcome the issues in these methods, an in-depth discussion on the state-of-the-art methods is required and thus, presented here. In this survey, we presented a detailed discussion of the existing methods to handle class overlap in imbalanced datasets with their advantages, disadvantages, limitations, and key performance metrics in which the method shown outperformed. The detailed comparative analysis mainly of recent years’ papers discussed and summarized the research gaps and future directions for the researchers in ML, DL, and BD-based applications.

Lee D., Kim J., Song S., Kim K.

Discovering sustainable business partnerships is crucial for small and medium-sized companies, where they can realize potential value through operational resources and abilities. Prior studies have mostly focused on predicting and developing new business partners using various machine learning techniques or social network analyses. However, effectively estimating potential benefits from business partnerships is much more valuable to companies. Therefore, this study proposes a method which combines deep learning and network analyses to estimate the potential value of business partnerships for companies. To demonstrate the effectiveness of the proposed method, we expand business partnerships between companies and assess potential value derived from the parenthesis using business transaction data collected from the Republic of Korea. The results suggest that companies can gain more potential value from extended networks when compared to previous ones. Furthermore, potential value results show clear distinctions between industries. Our findings provide evidence that small and medium-sized companies can experience significant benefits by establishing adequate business partnerships.

Li C., Ishak I., Ibrahim H., Zolkepli M., Sidi F., Li C.

Rinaldi, Ferdiana R., Setiawan N.A.