Deep Learning in Predicting Real Estate Property Prices: A Comparative Study

2023 | HICSS | Citations: 0

Authors: Shi, D.; Zhang, H.; Guan, J.; Zurada, J.; Chen, Z.; Li, X.

Abstract: The dominant methods for real estate property price prediction or valuation are ... Expand

Abstract: The dominant methods for real estate property price prediction or valuation are multi-regression based. Regression-based methods are, however, imperfect because they suffer from issues such as multicollinearity and heteroscedasticity. Recent years have witnessed the use of machine learning methods but the results are mixed. This paper introduces the application of a new approach using deep learning models to real estate property price prediction. The paper uses a deep learning approach for modeling to improve the accuracy of real estate property price prediction with data representing sales transactions in a large metropolitan area. Three deep learning models, LSTM, GRU and Transformer, are created and compared with other machine learning and traditional models. The results obtained for the data set with all features clearly show that the RF and Transformer models outperformed the other models. LSTM and GRU models produced the worst results, suggesting that they are perhaps not suitable to predict the real estate price. Furthermore, the implementations of Transformer and RF on a data set with feature reduction produced even more accurate prediction results. In conclusion, our research shows that the performance of the Transformer model is close to the RF model. Both models produce significantly better prediction results than existing approaches in terms of accuracy. Collapse

Topics: Python open source

Methods: deep learning transformer model machine learning autoregressive moving average model experiment

Responding to Online Reviews in Competitive Markets: A Controlled Diffusion Approach

2023 | Management Information Systems Quarterly | Citations: 0

Authors: Yang, Mingwen; Zheng, Zhiqiang (Eric); Mookerjee, Vijay; Chen, Hongyu

Abstract: We study how firms respond to online customer reviews in a competitive market w ... Expand

Abstract: We study how firms respond to online customer reviews in a competitive market where they jostle with one another for sales based on online ratings. The focus of this paper is on how firms can optimally manage their ratings through management response and how review ratings affect the sales and profits of competing firms. We develop a controlled diffusion process to model the coevolution of sales and ratings as a function of the response strategy chosen to maximize profit over time. Our model considers a variety of factors, such as profit margin and customer rating sensitivity, that influence a firm's effort to manage ratings and subsequently its sales and profits. More response effort needs to be exerted to manage ratings when either the profit margin of a tour is very high or customers are very sensitive to ratings. We estimate our model using data on Ctrip's tours that include each tour's sales, reviews, prices, and tour features. We find that consumers anchor their beliefs in the mean market rating and that their purchase decisions depend on the tour's rating relative to this anchor. Thus, relative, rather than absolute, ratings matter. Our study informs firms on how competition and other primitives impact their efforts to manage ratings and hence profit. Our methodology allowed us to conduct "what-if" analyses, for example, to study what would happen to the review ratings, sales, and profits of a tour if a firm adopted a different response strategy. We were also able to provide turnaround strategies for struggling tours, i.e., factors that a loss-making tour should change if it wishes to make a positive profit. Ultimately, we conducted a competitive analysis that allowed us to modify certain parameters that affect the intensity of competition and hence the sales and the profits of competing tours. Finally, we demonstrate the flexibility of the model by extending it to incorporate multiple state variables that might affect the response strategy. Collapse

Topics: application programming interface online learning system Twitter innovation management service quality

Methods: longitudinal research time series analysis generalized autoregressive conditional heteroskedasticity model natural language processing autoregressive moving average model

Theories: co-evolution theory

The Role and Value of Service Orchestration in Smart Grid Prosumer Service Systems

2022 | European Conference On Information Systems | Citations: 0

Authors: Merfeld, Tanja; Meisel, Stephan

Abstract: The implementation of smart grid infrastructure as well as the rise of eco-cons ... Expand

Abstract: The implementation of smart grid infrastructure as well as the rise of eco-conscious prosumers in the energy markets are leading to a paradigm shift in the energy sector. Residential households can no longer be viewed as passive market entities, but have to be considered as actors participating in value creation. In this work, we present the co-creation of value in energy markets through the lens of service dominant logic, and highlight the importance of service orchestrators for deriving both design decisions and operational decisions for complex energy systems. For the example of a real-time energy trading service, we assess the value of service orchestration by means of a simulation study. Thereby, we highlight the importance of service orchestration for creating valuable business opportunities, and we provide a transdisciplinary approach that combines service science and service operations research. Collapse

Topics: energy information system price management IT productivity decision making participatory design

Methods: simulation operations research experiment autoregressive moving average model design methodology

Theories: service-dominant logic

How Can Our Tweets Go Viral? Point-Process Modelling of Brand Content

2022 | Information & Management | Citations: 0

Authors: Zadeh, Amir; Sharda, Ramesh

Abstract: People create and share content via online social networks, which provide an unp ... Expand

Abstract: People create and share content via online social networks, which provide an unparalleled opportunity for brands to gain visibility, promote products or services and drive revenue growth. Much research has focused on why, how, or what social content is popular, trending and “hype”. One central challenge is to forecast the spread (cascades) of information that leads to the popularity of content throughout a social network. Online content tends to have bursts and spikes, experiencing a different cascading pattern depending on the viral propagation. In this paper, we propose and test a flexible framework capable of modelling such patterns and trends. We take temporal and network perspectives and develop a model based on the multivariate Hawkes processes that ac­ count for social behaviour and network elements such as follower counts, and activity variation observed in collective re-sharing behaviour. We focus on Twitter as the most widely used micro-blogging online social network and measure the popularity of a brand’s tweet by analysing the time-series path of three types of subsequent activities (retweets (RTs), replies (REs) and likes (LKs)). The specific model that we propose in this paper is the multidimensional epidemic-type aftershock sequence (METAS) model, a particular case of the multivariate Hawkes process. It consists of a power-law relaxation governing the timing of activities. It also includes an exponential boost as a reinforcement mechanism for the response amplitude to model the impact of influential users on their followers. Earlier attempts to model online cascades have treated all online responses as one type of activity. Rather than aggregating all the activities into one stream, and therefore, ignoring exciting effects among different types of activities, we incorporate the activity variation into the predictive models of content popularity, explicitly accounting for such excitation effects. We develop epidemic-type mutually exciting Hawkes point processes models to quantify such effects and to predict more accurately the number of follow-up activities (i.e., RTs, REs and LKs) on a brand tweet after it is posted. Our results suggest that the proposed model outperforms the state-of-the-art models in terms of prediction accuracy, as it is able to account for mutual ex­ citations and cross-interactions between sequences of users’ activities from one type to another. These results are relevant for developing and executing a plan for online activities by the brand owners. Collapse

Topics: Twitter social media social network pandemic marketing management

Methods: business process modeling longitudinal research time series analysis parametric test mathematical model

A Proposed Sentiment Analysis Deep Learning Algorithm for Analyzing COVID-19 Tweets

2021 | Information Systems Frontiers | Citations: 0

Authors: Kaur, Harleen; Ahsaan, Shafqat Ul; Alankar, Bhavya; Chang, Victor

Abstract: With the rise in cases of COVID-19, a bizarre situation of pressure was mounted ... Expand

Abstract: With the rise in cases of COVID-19, a bizarre situation of pressure was mounted on each country to make arrangements to control the population and utilize the available resources appropriately. The swiftly rising of positive cases globally created panic, anxiety and depression among people. The effect of this deadly disease was found to be directly proportional to the physical and mental health of the population. As of 28 October 2020, more than 40 million people are tested positive and more than 1 million deaths have been recorded. The most dominant tool that disturbed human life during this time is social media. The tweets regarding COVID-19, whether it was a number of positive cases or deaths, induced a wave of fear and anxiety among people living in different parts of the world. Nobody can deny the truth that social media is everywhere and everybody is connected with it directly or indirectly. This offers an opportunity for researchers and data scientists to access the data for academic and research use. The social media data contains many data that relate to real-life events like COVID-19. In this paper, an analysis of Twitter data has been done through the R programming language. We have collected the Twitter data based on hashtag keywords, including COVID-19, coronavirus, deaths, new case, recovered. In this study, we have designed an algorithm called Hybrid Heterogeneous Support Vector Machine (H-SVM) and performed the sentiment classification and classified them positive, negative and neutral sentiment scores. We have also compared the performance of the proposed algorithm on certain parameters like precision, recall, F1 score and accuracy with Recurrent Neural Network (RNN) and Support Vector Machine (SVM). Collapse

Topics: Twitter covid 19 pandemic programming language social media

Methods: support vector machine sentiment analysis recurrent neural network computational algorithm machine learning

Real-time temperature prediction in a cold supply chain based on Newton's law of cooling

2021 | Decision Support Systems | Citations: 0

Authors: Konovalenko, Iurii; Ludwig, André; Leopold, Henrik

Abstract: Many goods, including pharmaceuticals, require close temperature monitoring. Thi ... Expand

Abstract: Many goods, including pharmaceuticals, require close temperature monitoring. This is important not only for complying with regulations but also for guaranteeing safety of use. A particular challenge in controlling a product’s temperature arises during transportation. In cold supply chains (SCs), temperature is maintained by refrigerated containers. However, many situations, e.g. cooling system failure, lead to ambient temperature changes, and this needs to be detected as early as possible to prevent product damage. Existing approaches to temperature prediction are confined to long-term forecasts with relatively stable ambient temperatures and/or rely on multiple sensors in the known fixed positions. Since interventions in a SC are required immediately, there is a need for methods that provide real-time predictions regarding regular ambient temperature instability, i.e. when the ambient temperature changes unexpectedly in the short term. We propose a novel method that extends the applicability of Newton’s law of cooling (NLC) to changeable ambient temperatures based on a set of tem­ perature stability conditions and a sensor measurement error. In the method, an optimal number of measure­ ments that characterize stable ambient temperatures and improve prediction reliability are selected. We compare the adapted NLC with artificial neural networks and autoregressive moving average models with respect to deviation prediction, prediction error, and execution time. Our evaluation based on real-world data shows that the adapted NLC outperforms existing baseline methods. In contrast to existing solutions, our method does not require any knowledge about the positioning of products within the container, further increasing its practical value. Collapse

Topics: Python logistics management open source

Methods: autoregressive moving average model experiment longitudinal research machine learning artificial neural network

First, Do No Harm: Predictive Analytics to Reduce In-Hospital Adverse Events

2021 | Journal of Management Information Systems | Citations: 0

Authors: Lin, Yu-Kai; Fang, Xiao

Abstract: Inadequate patient safety is a serious issue in current medical practice. Medica ... Expand

Abstract: Inadequate patient safety is a serious issue in current medical practice. Medical errors cause adverse events (AEs) for patients and lead to premature deaths, unintended complications, prolonged hospital stays, and higher medical costs. Although the importance of AE prediction and prevention is well recognized in the information systems literature, there is a dearth of research on modeling and predicting AEs caused by medical errors. Following the design science research paradigm, this study describes the search, design, and evaluation of a novel in-hospital AE prediction model, called Stochastic Autoregressions for Latent Trajectories (SALT). The proposed model uniquely integrates generalized linear mixed model with multitask learning and stochastic time-series processes. Results from our empirical evaluation show that SALT outperforms prior state-of-the-art techniques in predicting AEs during patients' hospital stays. Through a simulation, we further demonstrate significant cost savings potential when hospitals implement and integrate SALT in their inpatient care. This study contributes to the design science literature by formalizing the in-hospital AE prediction problem, on the one hand, and developing a novel graphical model to address the prediction problem, on the other. For healthcare practitioners and administrators, our predictive analytics approach unveils important insights to minimize AEs. Collapse

Topics: organizational value knowledge base anonymity health information system problem solving

Methods: machine learning longitudinal research feature engineering time series analysis experiment

New algorithms for automatic modelling and forecasting of decision support systems

2021 | Decision Support Systems | Citations: 2

Authors: Pedregal, Diego J.

Abstract: Decision support systems often rely on time series forecasting, making the accur ... Expand

Abstract: Decision support systems often rely on time series forecasting, making the accuracy of such systems of paramount importance for their efficiency. Since most systems nowadays require the processing of massive amount of data, automatic identification of time series models has become inevitable. This automatism is inherent in artificial intelligence methods, but it often goes unnoticed that forecasting ‘classical’ methods have also been developing their own automatic methods for a long time. The radical novelty of this paper is the development of a brand new algorithm for identification of structural Unobserved Components models from which decision support systems may benefit. A second point is that combination of forecasts is more fruitful than competition or method selection in some cases. Both points are illustrated in two examples that show the effectiveness of the identification procedure and the forecasting gains when fairly different methods are combined. Collapse

Topics: decision support system

Methods: computational algorithm autoregressive integrated moving average time series analysis longitudinal research Student's t-test

Model identification for ARMA time series through convolutional neural networks

2021 | Decision Support Systems | Citations: 0

Authors: Tang, Wai Hoh; Röllin, Adrian

Abstract: We use convolutional neural networks for model identification in ARMA time serie ... Expand

Abstract: We use convolutional neural networks for model identification in ARMA time series models, where our networks are trained on synthetic data with known ground truths. Comparing the performance of these networks with traditional likelihood-based methods, in particular the Akaike and Bayesian Information Criteria, we are able to show that when it comes to statistical inference on ARMA orders, neural networks can significantly outperform likelihood-based methods in terms of accuracy and, by orders of magnitude, in terms of speed. We also observe improvements in terms of time series forecasting. Our approach shows the feasibility of using artificial neural networks for statistical inference in situations where classical likelihood-based methods are difficult or costly to implement. Collapse

Methods: time series analysis longitudinal research autoregressive moving average model artificial neural network residual neural network

Theories: mathematical theory

Relational Time Series Forecasting for Retail Drugstores: A Graph Neural Network Approach

2021 | International Conference on Information Systems | Citations: 0

Authors: Liu, Jing; Wang, Gang; Huang, Lihua

Abstract: Forecasting retail sales accurately is important for supporting strategic, tact ... Expand

Abstract: Forecasting retail sales accurately is important for supporting strategic, tactic, and operational decisions. Interdependencies among groups of time series are ubiquitous in real scenarios. However, previous research either completely ignores or fails to exploit these cross-series dependencies effectively and efficiently. This study follows the design science paradigm, and develops an innovative deep learning framework for long-term sales forecasting of retail drugstores. A novel graph neural network module is designed and developed for relational time series forecasting to capture cross-series (i.e., global) patterns. Moreover, we design a multi-source fusion module to fuse global patterns, specific temporal patterns of each time series, and context features. Experimental results demonstrate the effectiveness of our approach for both next-step and multi-step forecasting, and verify the utility of several crucial components in the framework. The proposed approach has important practical implications and research contributions. Collapse

Topics: geographic information system Python application programming interface information exchange

Methods: longitudinal research time series analysis experiment ablation analysis autoregressive moving average model