Search results

The purpose of this study is to develop a review rating prediction method based on a supervised text mining approach for unrated customer reviews.

Using 2,851 hotel comment card (HCC) reviews, this paper manually labeled positive and negative comments with seven aspects (dining, cleanliness, service, entertainment, price, public, room) that emerged from the content of said reviews. After text preprocessing (tokenization, eliminating punctuation, stemming, etc.), two classifier models were created for predicting the reviews’ sentiments and aspects. Thus, an aggregate rating scale was generated using these two classifier models to determine overall rating values.

A new algorithm, Comment Rate (CRate), based on supervised learning, is proposed. The results are compared with another review-rating algorithm called location based social matrix factorization (LBSMF) to check the consistency of the proposed algorithm. It is seen that the proposed algorithm can predict the sentiments better than LBSMF. The performance evaluation is performed on a real data set, and the results indicate that the CRate algorithm truly predicts the overall rating with ratio 80.27%. In addition, the CRate algorithm can generate an overall rating prediction scale for hotel management to automatically analyze customer reviews and understand the sentiment thereof.

The review data were only collected from a resort hotel during a limited period. Therefore, this paper cannot explore the effect of independent variables on the dependent variable in context of larger period.

This paper provides a novel overall rating prediction technique allowing hotel management to improve their operations. With this feature, hotel management can evaluate guest feedback through HCCs more effectively and quickly. In this way, the hotel management will be able to identify those service areas that need to be developed faster and more effectively. In addition, this review rating prediction approach can be applied to customer reviews posted via online platforms for detecting review and rating reliability.

Manually analyzing textual information is time-consuming and can lead to measurement errors. Therefore, the primary contribution of this study is that although comment cards do not have rating values, the proposed CRate algorithm can predict the overall rating and understand the sentiment of the reviews in question.

This paper aims to propose a trust-based ant-colony recommender system. It achieves high accuracy and coverage by integrating the importance level of friends. This paper has two main contributions, namely, selecting higher-quality raters and improving the prediction step. From these two contributions, the proposed trust-based ant-colony recommender system could provide more accurate and wider-coverage prediction than existing systems.

To obtain higher-quality raters, the data set was preprocessed, and then, trust values were calculated. The depth of search was increased to obtain higher coverage levels. This work also focuses on the importance level of friends in the system. Because the levels of influence on the active user of all friends are not equal, the importance level of friends is integrated into the system by transposing rater’s rating to the active user’s perspective and then assigning a weight to each rater.

The experimental evaluation clearly demonstrates that the proposed method achieves better results in terms of both accuracy and coverage than existing trust-based recommender systems. It was found that integrating the importance level of friends into the system, which transposes ratings and assigns weight to each user, can increase accuracy and coverage.

Existing trust-based ant-colony recommender systems do not consider the importance level of friends in the prediction step. Most of them only focus on finding raters and then using the rater’s real ratings in the prediction step. A new method is proposed that integrates the importance level of friends into the system by transposing a rater’s rating to match the active user’s perspective and assigning a weight for each rater. The experimental evaluation demonstrates that the proposed method achieves better accuracy and coverage than existing systems.

Successful bridge management system (BMS) development requires a reliable bridge deterioration model, which is the most crucial component in a BMS. Historical condition ratings obtained from biennial bridge inspections are a major source for predicting future bridge deterioration in BMSs. However, historical condition ratings are very limited in most bridge agencies, thus posing a major barrier for predicting reliable future bridge performance. The purpose of this paper is to present a preliminary study as part of a long‐term research on the development of a reliable bridge deterioration model using advanced Artificial Intelligence (AI) techniques.

This proposed study aims to develop a reliable deterioration model. The development work consists of two major Stages: stage 1 – generating unavailable bridge element condition rating records using the Backward Prediction Model (BPM). This helps to provide sufficient historical deterioration patterns for each element; and stage 2 – predicting long‐term condition ratings based on the outcome of Stage 1 using time delay neural networks (TDNNs).

Long‐term prediction using proposed method can also be expressed in the same form of inspection records – element quantities of each bridge element can be predicted. The proposed AI‐based deterioration model does not ignore critical failure risks in small number of bridge elements in low condition states (CSs). This implies that the risk in long‐term predictions can be reduced.

The proposed methodology aims to utilise limited bridge inspection records over a short period to predict large datasets spanning over a much longer time period for a reliable, accurate and efficient long‐term bridge deterioration model. Typical uncertainty, due to the limitation of overall condition rating (OCR) method, can be minimised in long‐term predictions using limited inspection records.

Collaborative filtering (CF), one of the most popular recommendation techniques, is based on the principle of word-of-mouth communication between other like-minded users. The process of identifying these like-minded or similar users remains crucial for a CF framework. Conventionally, a neighbor is the one among the similar users who has rated the item under consideration. To select neighbors by the existing practices, their similarity deteriorates as many similar users might not have rated the item under consideration. This paper aims to address the drawback in the existing CF method where “not-so-similar” or “weak” neighbors are selected.

The new approach proposed here selects neighbors only on the basis of highest similarity coefficient, irrespective of rating the item under consideration. Further, to predict missing ratings by some neighbors for the item under consideration, ordinal logistic regression based on item–item similarity is used here.

Experiments using the MovieLens (ml-100) data set prove the efficacy of the proposed approach on different performance evaluation metrics such as accuracy and classification metrics. Apart from higher prediction quality, coverage values are also at par with the literature.

This new approach gets its motivation from the principle of the CF method to rely on the opinion of the closest neighbors, which seems more meaningful than trusting “not-so-similar” or “weak” neighbors. The static nature of the neighborhood addresses the scalability issue of CF. Use of ordinal logistic regression as a prediction technique addresses the statistical inappropriateness of other linear models to make predictions for ordinal scale ratings data.

Artificial intelligence (AI) has started to receive attention in the field of digital entrepreneurship. However, few studies propose AI-based models aimed at assisting entrepreneurs in their day-to-day operations. In addition, extant models from the product design literature, while technically promising, fail to propose methods suitable for opportunity development with high level of uncertainty. This study develops and tests a predictive model that provides entrepreneurs with a digital infrastructure for automated testing. Such an approach aims at harnessing AI-based predictive technologies while keeping the ability to respond to the unexpected.

Based on effectuation theory, this study identifies an AI-based, predictive phase in the “build-measure-learn” loop of Lean startup. The predictive component, based on recommendation algorithm techniques, is integrated into a framework that considers both prediction (causal) and controlled (effectual) logics of action. The performance of the so-called active learning build-measure-predict-learn algorithm is evaluated on a data set collected from a case study.

The results show that the algorithm can predict the desirability level of newly implemented product design decisions (PDDs) in the context of a digital product. The main advantages, in addition to the prediction performance, are the ability to detect cases where predictions are likely to be less precise and an easy-to-assess indicator for product design desirability. The model is found to deal with uncertainty in a threefold way: epistemological expansion through accelerated data gathering, ontological reduction of uncertainty by revealing prior “unknown unknowns” and methodological scaffolding, as the framework accommodates both predictive (causal) and controlled (effectual) practices.

Research about using AI in entrepreneurship is still in a nascent stage. This paper can serve as a starting point for new research on predictive techniques and AI-based infrastructures aiming to support digital entrepreneurs in their day-to-day operations. This work can also encourage theoretical developments, building on effectuation and causation, to better understand Lean startup practices, especially when supported by digital infrastructures accelerating the entrepreneurial process.

To test if the factors “reviewer location” and “time frame” have any impact on the prediction results when predicting online product ratings from user reviews.

Reviews and ratings are scraped for the product “The Secret” book through Web pages of e-commerce websites like Amazon and Flipkart. Such data is used for training the model to predict ratings of similar products based on reviews data in various other social media platforms like Facebook, Quora and YouTube. After data preprocessing, sentiment analysis is used for opinion classification. A multi-class supervised support vector machine is used for feature classification and predictions. The four models produced in the study have a prediction accuracy of 79%. The data collection is done based on a specific geographical location and specific time frame. Post evaluating the predictions, inferential statistics are used to check for significance.

There will be an impact on the ratings predicted from the reviews that belong to a particular geographic location or time frame. The ratings predicted from such reviews help in taking accurate decisions as they are robust and informative.

This study is confined to a single product and for cross domain social media pages, only Facebook, YouTube and Quora data are considered.

Provides credible ratings of a product/service on all cross domain social media pages making the initial screening process of purchase decisions better.

Many studies explored the usefulness of reviews for rating prediction based on review nature. This study aims to identify the usefulness of reviews based on factors that would reduce uncertainty in the purchase process.

Despite the growing importance of online word-of-mouth (WOM) with regard to television (TV) ratings, it is usually excluded from early prediction models. The purpose of this paper is to investigate the role of online WOM in TV ratings predictions, focussing on whether the incorporation of online WOM could improve predictions of TV ratings, and extracts meaningful rules for decision-making.

The author uses online WOM as a potential predictive variable in the TV ratings prediction model. The author matches a list of programs based on TV ratings for the movie channel with internet user reviews and TV ratings information from Yahoo! Movies (YM) and XYZ Company. The data set includes 71 movies, for which the data were analyzed with a hybrid model.

Grey relational analysis shows that online WOM is a useful ex ante determinant of TV ratings. As a predictive variable, it plays an essential role in enhancing TV ratings predictions. The experimental results also indicate that the proposed model surpasses other listed methods in terms of both accuracy and reduction of variables, while the proposed procedure yields a set of easily understandable decision rules that facilitate the interpretation of TV ratings information.

This paper identifies critical predictors of TV ratings and suggests that online WOM messages are a credible source. A hybrid model is developed to illustrate an intelligent prediction system for TV ratings.

The study demonstrates the effectiveness of online WOM and its impact on TV ratings. It offers an intelligent prediction system for TV ratings with practical implications for managers within the TV industry.

This study attempts to identify fundamental determinants of bond ratings for non-financial and financial firms. Further the study aims to develop a parsimonious bond rating model and compare its efficacy across statistical and range of machine learning methods in the Indian context. The study is motivated by the insufficiency of prior work in the Indian context.

The authors identify the critical determinants of non-financial and financial firms using multinomial logistic regression. Various machine learning and statistical methods are employed to identify the optimal bond rating prediction model. The data cover 8,346 bond issues from 2009 to 2019.

The authors find that industry concentration, sales, operating leverage, operating efficiency, profitability, solvency, strategic ownership, age, firm size and firm value play an important role in rating non-financial firms. Operating efficiency, profitability, strategic ownership and size are also relevant for financial firms besides additional determinants related to the capital adequacy, asset quality, management efficiency, earnings quality and liquidity (CAMEL) approach. The authors find that random forest outperforms logit and other machine learning methods with an accuracy rate of 92 and 91% for non-financial and financial firms.

The study identifies important determinants of bond ratings for both non-financial and financial firms. The study interalia finds that the random forest technique is the most appropriate method for bond ratings predictions in India.

Better bond ratings may mitigate corporate defaults.

Unlike prior literature, the study identifies determinants of bond ratings for both non-financial and financial firms. The study also experiments with modern machine learning techniques besides the traditional statistical approach for model building in case of relatively under researched market.

Given the importance of credit rating agencies’ (CRAs) assessment in affecting international financial markets, it is useful for policymakers and investors to be able to forecast it properly. Therefore, this study aims to forecast sovereign risk perception of the main agencies related to Brazilian bonds through the application of different machine learning (ML) techniques and evaluate their predictive accuracy in order to find out which one is best for this task.

Based on monthly data from January 1996 to November 2018, we perform different forecast analyses using the K-Nearest Neighbors, the Gradient Boosted Random Trees and the Multilayer Perceptron methods.

The results of this study suggest the Multilayer Perceptron technique is the most reliable one. Its predictive accuracy is relatively high if compared to the other two methods. Its forecast errors are the lowest in both the out-of-sample and in-sample forecasts’ exercises. These results hold if we consider the CRAs classification structure as linear or logarithmic. Moreover, its forecast errors are not statistically associated with periods of changes in CRAs’ opinion of any sort.

To the best of the authors’ knowledge, this study is the first to evaluate the performance of ML methods in the task of predicting sovereign credit news, including not only the sovereign ratings but also the outlook and credit watch status. In addition, the authors investigate whether the forecasts errors are statistically associated with periods of changes in sovereign risk perception.

In recent years, a sizeable amount of research in finance and accounting has been devoted to the issue of bond rating and bond rating changes. A major thrust of these research efforts was to develop and test some prediction‐based models using mainly financial ratios and their trends. This paper tests the ability of statistical decomposition analysis of financial statements to predict bond rating changes. The results show that the decomposition analysis almost does not beat the a priori probability model and is no better than multiple discriminant analysis using simple financial ratios. One important piece of information for participants in debt markets is the assessment of the relative risk associated with a particular bond issue, commonly known as bond ratings. These ratings, however, are not usually fixed for the life of the issues. From time to time, the rating agencies review their ratings of the outstanding bond issues and make changes to these ratings (either upward or downward) when needed. Over the years, researchers have attempted to develop and test some prediction based models in order to predict bond ratings or bond rating changes. These prediction models have employed some variables that are assumed to reflect the rating agency decision‐making activities. Although the rating process is complicated and based mainly on judgmental considerations, Hawkins, Brown and Campbell (1983, p. 95) reported that the academic research strongly suggests that a reliable estimate of a potential bond rating or rating change can be determined by a few key financial ratios. Information theory decomposition measures have received in recent years considerable attention as a potential tool for predicting corporate events, namely corporate bankruptcy (e.g., Lev 1970; Moyer 1977; Walker, Stowe and Moriarity 1979; Booth 1983). The underlying proposition in these studies is that corporate failure, as an event, is expected to be preceded by significant changes in the company's assets and liabilities structure. Although the event of bond rating changes is different from the bankruptcy event in terms of consequences, one can still propose that a bond rating change, as a corporate event, is also expected to be preceded by some significant changes in the company's assets and liabilities structure. Therefore, the decomposition analysis may have a predictive ability in the case of bond rating changes. The purpose of this paper is to empirically test and compare the classification and predictive accuracy of the decomposition analysis with the performance of a multiple discriminant model that uses financial ratios and their trends in the context of bond rating changes.