Search

In this paper we present C o R o NN a a deep sequential framework for epidemic prediction that leverages a flexible combination of sequential and convolutional components to analyse the transmission of COVID-19 and, perhaps, other undiscovered viruses. Importantly, our methodology is generic and may be tailored to specific analysis goals. We exploit C o R o NN a to analyse the impact of various mobility containment policies on the pandemic using cumulative viral dissemination statistics with local demographic and movement data from several nations. Our experiments show that data on mobility has a significant, but delayed, impact on viral propagation. When compared to alternative frameworks that combine multivariate lagged predictors and basic LSTM models, C o R o NN a outperforms them. On the contrary, no technique based solely on lagged viral dissemination statistics can forecast daily cases.

The accurate prediction of human movement trajectory has a variety of benefits for many applications such as optimizing nurse’s trajectory in a hospital, optimizing movements of old or disabled people to minimize their routine efforts, etc. To perform human movement prediction, large amount of historical positioning data from sensors has to be collected and mined. We analyzed different human sequential movement prediction approaches and their limitations. In this work, we propose a new classifier named Apriori based Probability Tree Classifier (APTC) which predicts the human movement sequence patterns in indoor environment. The APTC classifier is integrated into Bagged J48 Machine learning algorithm which results in an ensemble model to predict the future human movement patterns. The patterns are mined based on spatial, temporal and social data which add more accuracy to our prediction. Our model also performs clustering mechanism to detect the abnormal patterns.

Hard interaction learning between source sequences and their next targets is challenging, which exists in a myriad of sequential prediction tasks. During the training process, most existing methods focus on explicitly hard interactions caused by wrong responses. However, a model might conduct correct responses by capturing a subset of learnable patterns, which results in implicitly hard interactions with some unlearned patterns. As such, its generalization performance is weakened. The problem gets more serious in sequential prediction due to the interference of substantial similar candidate targets. To this end, we propose a Hardness Aware Interaction Learning framework (HAIL) that mainly consists of two base sequential learning networks and mutual exclusivity distillation (MED). The base networks are initialized differently to learn distinctive view patterns, thus gaining different training experiences. The experiences in the form of the unlikelihood of correct responses are drawn from each other by MED, which provides mutual exclusivity knowledge to figure out implicitly hard interactions. Moreover, we deduce that the unlikelihood essentially introduces additional gradients to push the pattern learning of correct responses. Our framework can be easily extended to more peer base networks. Evaluation is conducted on four datasets covering cyber and physical spaces. The experimental results demonstrate that our framework outperforms several state-of-the-art methods in terms of top-k based metrics., 10 pages; 4 figures; Proceedings of the 30th ACM International Conference on Information and Knowledge Management (CIKM '21), November 1--5, 2021, Virtual Event, QLD, Australia

In this article, we propose a hybrid-Bayesian frequentist approach using a Bayesian sequential prediction of the index of satisfaction. For interim analysis that addresses prediction hypoth...

Predicting a click sequence in mobile applications improves the user experience in various ways. By predicting which button will be clicked next, one can predict how the application will work and how the device will operate. However, predicting the click sequence is difficult because of the problems involved in collecting click sequences in real application usage. More importantly, accurate predictions are extremely challenging. In this paper, we address these issues. We propose PathFinder, a scheme for collecting click events and based on them predicting the next click in the application. The clicks are collected with the Android Accessibility Service and the next click is predicted via long short-term memory (LSTM). For the prediction, the base click sequence model is first generated from all users’ data; then, a personalized model is trained with an individual click sequence. As training data considerably influences the performance of LSTM, several techniques are developed to enhance the quality of the training data. The experimental results for 100 popular applications showed that the coverage and accuracy of click sequence tracing were 95% and 96%, respectively. Furthermore, PathFinder predicted the top three buttons that would be clicked next with a 0.76 F -measure for 1 775 043 real click data.

The generalization of predictive models that follow the standard risk minimization paradigm of machine learning can be hindered by the presence of spurious correlations in the data. Identifying invariant predictors while training on data from multiple environments can influence models to focus on features that have an invariant causal relationship with the target, while reducing the effect of spurious features. Such invariant risk minimization approaches heavily rely on clearly defined environments and data being perfectly segmented into these environments for training. However, in real-world settings, perfect segmentation is challenging to achieve and these environment-aware approaches prove to be sensitive to segmentation errors. In this work, we present an environment-agnostic approach to develop generalizable models for classification tasks in sequential datasets without needing prior knowledge of environments. We show that our approach results in models that can generalize to out-of-distribution data and are not influenced by spurious correlations. We evaluate our approach on real-world sequential datasets from various domains.

Sequence prediction problems have been a major problem for a long time. Recurrent Neural Network (RNN) has been a good solution for sequential prediction problems. This work aims to create a generative model for text. Even though, RNN has its own limitations such as vanishing and exploding gradient descent problems, and inefficiency to keep track of long-term dependencies. To overcome these drawbacks, Long Short Term Memory (LSTM) has been a path-breaking solution to deal with sequential data and text data in particular. This paper delineates the design and working of text generation using word-level LSTM-RNN.

Clinical event sequences consist of thousands of clinical events that represent records of patient care in time. Developing accurate prediction models for such sequences is of a great importance for defining representations of a patient state and for improving patient care. One important challenge of learning a good predictive model of clinical sequences is patient-specific variability. Based on underlying clinical complications, each patient’s sequence may consist of different sets of clinical events. However, population-based models learned from such sequences may not accurately predict patient-specific dynamics of event sequences. To address the problem, we develop a new adaptive event sequence prediction framework that learns to adjust its prediction for individual patients through an online model update.

Many online fighting games use rollback netcode in order to compensate for network delay. Rollback netcode allows players to experience the game as having reduced delay. A drawback of this is that players will sometimes see the game quickly ”jump” to a different state to adjust for the the remote player’s actions. Rollback netcode implementations require a method for predicting the remote player’s next button inputs. Current implementations use a naive repeatlastframe policy for such prediction. There is a possibility that alternative methods may lead to improved user experience. This project examines the problem of improving input prediction in fighting games. It details the development of a new prediction model based on recurrent neural networks. The model was trained and evaluated using a dataset of several thousand recorded player input sequences. The results show that the new model slightly outperforms the naive method in prediction accuracy, with the difference being greater for longer predictions. However, it has far higher requirements both in terms of memory and computation cost. It seems unlikely that the model would significantly improve on current rollback netcode implementations. However, there may be ways to improve predictions further, and the effects on user experience remains unknown. Många online fightingspel använder rollback netcode för att kompensera för nätverksfördröjning. Rollback netcode låter spelare uppleva spelet med mindre fördröjning. En nackdel av detta är att spelare ibland ser spelet snabbt ”hoppa” till ett annat tillstånd för att justera för motspelarens handlingar. Rollback netcode implementationer behöver en policy för att förutsäga motspelarens nästa knapptryckningar. Nuvarande implementationer använder en naiv repetera-senaste-frame policy för förutsägelser. Det finns en möjlighet att alternativa metoder kan leda till förbättrad användarupplevelse. Det här projektet undersöker problemet att förbättra förutsägelser av knapptryckningar i fightingspel. Det beskriver utvecklingen av en ny förutsägelsemodell baserad på rekursiva neuronnät. Modellen tränades och evaluerades med ett dataset av flera tusen inspelade knappsekvenser. Resultaten visar att den nya modellen överträffar den naiva metoden i noggrannhet, med större skillnad för längre förutsägelser. Dock har den mycket högre krav i både minne och beräkningskostad. Det verkar osannolikt att modellen skulle avsevärt förbättra nuvarande rollback netcode implementationer. Men det kan finnas sätt att förbättra förutsägelser ytterligare, och påverkan på användarupplevelsen förblir okänd.

This paper presents an analysis of informative presentations using sequential multimodal modeling for automatic assessment of presentation performance. For this purpose, we transform a single video into multiple time-series segments that are provided as inputs to sequential models, such as Long Short-Term Memory (LSTM). This sequence modeling approach enables us to capture the time-series change of multimodal behaviors during the presentation. We proposed variants of sequential models that improve the accuracy of performance prediction over non-sequential models. Moreover, we performed segment analysis on the sequential models to analyze how relevant information from various segments can lead to better performance in sequential prediction models.

Human beings nowadays spend more than 90% of the lifetime indoors, leading to the dramatic increase of energy consumption in various buildings. Therefore, research regarding the environment friendly building becomes much more popular recently in which the prediction of energy consumption is a promised method. Nevertheless, the accuracy of prediction is not sound due to insufficient samples. A novel data generation model, termed HMSP, based on the generative adversarial networks, is proposed in this paper to generate much more data robustly, depending on a small number of samples available. The prediction CV-RMSE results, adopting data from the hybrid model, reach 3.03% at best and 7.99% at worst respectively compared to the samples recorded.

Type 2 diabetes mellitus (T2DM) has been identified as one of the most challenging chronic diseases to manage. In recent years, the incidence of T2DM has increased, which has seriously endangered people’s health and life quality. Glycosylated hemoglobin (HbA1c) is the gold standard clinical indicator of the progression of T2DM. An accurate prediction of HbA1c levels not only helps medical workers improve the accuracy of clinical decision-making but also helps patients to better understand the clinical progression of T2DM and conduct self-management to achieve the goal of controlling the progression of T2DM. Therefore, we introduced the long short-term memory (LSTM) neural network to predict patients’ HbA1c levels using time sequential data from electronic medical records (EMRs). We added the self-attention mechanism based on the traditional LSTM to capture the long-term interdependence of feature elements and which ensure that the memory was more profound and effective, and used the gradient search technology to minimize the mean square error of the predicted value of the network and the real value. LSTM with the self-attention mechanism performed better than the traditional deep learning sequence prediction method. Our research provides a good reference for the application of deep learning in the field of medical health management.

An automated smart home system is a key contributor to user assistance technology in modern civilization. Crucial merit of such a system is its ability to train itself through recorded data and recognize patterns in resident behaviors. Lack of sufficient prediction accuracy, exponential memory consumption, and extensive runtime prevent many of the current activity prediction approaches from being seamlessly integrated into consumer residences. This research introduces a sequence prediction algorithm which uses a prefix tree-based data model in order to learn and predict user actions. The algorithm applies episode discovery to detect correlated sensor events and learns the activities using a lossless data compression technique. This process assigns a probability of occurrence to sensor events and uses these probabilities to detect patterns in resident behavior. A complexity analysis of the algorithm is done to prove its efficiency in terms of memory usage and runtime. Using the presented technique, predictions are performed on popular datasets and contrasted with existing algorithms. The proposed algorithm achieves an 8.22% improvement in prediction accuracy over its predecessors, along with 66.69% better memory efficiency and 37% faster runtime.

A contemporary approach for acquiring the computational gains of depth in recurrent neural networks (RNNs) is to hierarchically stack multiple recurrent layers. However, such performance gains come with the cost of challenging optimization of hierarchal RNNs (HRNNs) which are deep both hierarchically and temporally. The researchers have exclusively highlighted the significance of using shortcuts for learning deep hierarchical representations and deep temporal dependencies. However, no significant efforts are made to unify these finding into a single framework for learning deep HRNNs. We propose residual recurrent highway network (R2HN) that contains highways within temporal structure of the network for unimpeded information propagation, thus alleviating gradient vanishing problem. The hierarchical structure learning is posed as residual learning framework to prevent performance degradation problem. The proposed R2HN contain significantly reduced data-dependent parameters as compared to related methods. The experiments on language modeling (LM) tasks have demonstrated that the proposed architecture leads to design effective models. On LM experiments with Penn TreeBank, the model achieved 60.3 perplexity and outperformed baseline and related models that we tested.

The ability to know in advance the trend of running process instances, with respect to different features, such as the expected completion time, would allow business managers to timely counteract to undesired situations, in order to prevent losses. Therefore, the ability to accurately predict future features of running business process instances would be a very helpful aid when managing processes, especially under service level agreement constraints. However, making such accurate forecasts is not easy: many factors may influence the predicted features. Many approaches have been proposed to cope with this problem but, generally, they assume that the underlying process is stationary. However, in real cases this assumption is not always true. In this work we present new methods for predicting the remaining time of running cases. In particular we propose a method, assuming process stationarity, which achieves state-of-the-art performances and two other methods which are able to make predictions even with non-stationary processes. We also describe an approach able to predict the full sequence of activities that a running case is going to take. All these methods are extensively evaluated on different real case studies.

While activity recognition has been shown to be valuable for pervasive computing applications, less work has focused on techniques for forecasting the future occurrence of activities. We present an activity forecasting method to predict the time that will elapse until a target activity occurs. This method generates an activity forecast using a regression tree classifier and offers an advantage over sequence prediction methods in that it can predict expected time until an activity occurs. We evaluate this algorithm on real-world smart home datasets and provide evidence that our proposed approach is most effective at predicting activity timings.

With an increasing demand for bandwidth and lower latency in mobile communication networks it becomes gradually more important to improve current mobile network management solutions using available network data. To improve the network management it can for instance be of interest to infer future available bandwidth to the end user of the network. This can be done by utilizing the current knowledge of real-time user equipment (UE) behaviour in the network. In the scope of this thesis interest lies in, given a set of visited radio access points (cells), to predict what the next one is going to be. For this reason the aim is to investigate the prediction performance when utilizing the All-K-Order Markov (AKOM) model, with some added variations, on collected data generated from train trajectories. Moreover a method for testing the suitability of modeling the sequence of cells as a time-homogeneous Markov chain is proposed, in order to determine the goodness-of- t with the available data. Lastly, the elapsed time in each cell is attempted to be predicted using linear regression given the prior history window of previous cell and elapsed times pairs. The results show that moderate to good prediction accuracy on the upcoming cell can be achieved with AKOM and associated variations. For predicting the upcoming sojourn time in future cells the results reveal that linear regression does not yield satisfactory results and possibly another regression model should be utilized. Med en ökande efterfrågan på banbredd och kortare latens i mobila nätverk har det gradvis blivit viktigare att förbättra nuvarande lösningar för hantering av nätverk genom att använda tillgänglig nätverksdata. Specifikt är det av intresse att kunna dra slutsatser kring vad framtida bandbredsförhållanden kommer vara, samt övriga parametrar av intresse genom att använda tillgänglig information om aktuell mobil användarutrustnings (UE) beteende i det mobila nätverket. Inom ramen av detta masterarbete ligger fokus på att, givet tidigare besökta radio accesspunkter (celler), kunna förutspå vilken nästkommande besökta cell kommer att vara. Av denna anledning är målet att undersöka vilken prestanda som kan uppnås när All-$K$-Order Markov (AKOM) modellen, med associerade varianter av denna, används på samlad data från tågfärder. Dessutom ges det förslag på test som avgör hur lämpligt det är att modelera observerade sekvenser av celler som en homogen Markovkedja med tillgänglig data. Slutligen undersöks även om besökstiden i en framtida cell kan förutspås med linjär regression givet ett historiskt fönster av tidigare cell och besökstids par. Erhållna resultat visar att måttlig till bra prestanda kan uppnås när kommande celler förutspås med AKOM modellen och associerade variationer. För prediktering av besökstid i kommande cell med linjär regression erhålles det däremot inte tillfredsställande resultat, vilket tyder på att en alternativ regressionsmetod antagligen är bättre lämpad för denna data.

Predicting users daily location visits - when and where they will go, and how long they will stay - is key for making effective location-based recommendations. Knowledge of an upcoming day allows the suggestion of relevant alternatives (e.g., a new coffee shop on the way to work) in advance, prior to a visit. This helps users make informed decisions and plan accordingly. People's visit routines, or just routines, can vary significantly from day to day, and visits from earlier in the day, week, or month may affect subsequent choices. Traditionally, routine prediction has been modeled with sequence methods, such as HMMs or more recently with RNN-based architectures. However, the problem with such architectures is that their predictive performance degrades when increasing the number of historical observations in the routine sequence. In this paper, we propose Masked-TCN (MTCN), a novel method based on time-dilated convolutional networks. The method implements custom dilations and masking which can process effectively long routine sequences, identifying recurring patterns at different resolution - hourly, daily, weekly, monthly. We demonstrate that MTCN achieves 8% improvement in accuracy over current state-of-the-art solutions on a large data set of visit routines.

Data sparsity is one of the most challenging problems in recommendation systems. In this paper, we tackle this problem by proposing a novel multi-modal behavioral information-aware recommendation method named MIAR which is based on recurrent neural networks and matrix factorization. First, an interaction context-aware sequential prediction model is designed to capture user-item interaction contextual information and behavioral sequence information. Second, an attributed context-aware rating prediction model is proposed to capture attribution contextual information and rating information. Finally, three fusion methods are developed to combine two sub-models. As a result, the MIAR method has several distinguished advantages in terms of mitigating the data sparsity problem. The method can well perceive diverse influences of interaction and attribution contextual information. Meanwhile, a large number of behavioral sequence and rating information can be utilized by the MIAR approach. The proposed algorithm is evaluated on real-world datasets and the experimental results show that MIAR can significantly improve recommendation performance compared to the existing state-of-art recommendation algorithms.

Long-term load forecast (LTLF) for area distribution feeders is one of the most critical tasks frequently performed in electric distribution utility companies. For a specific planning area, cost-effective system upgrades can only be planned out based on accurate feeder LTLF results. In our previous research, we established a unique sequence prediction method which has the tremendous advantage of combining area top-down, feeder bottom-up and multi-year historical data all together for forecast and achieved a superior performance over various traditional methods by real-world tests. However, the previous method only focused on the forecast of the next one-year. In our current work, we significantly improved this method: the forecast can now be extended to a multi-year forecast window in the future; unsupervised learning techniques are used to group feeders by their load composition features to improve accuracy; we also propose a novel selective sequence learning mechanism which uses Gated Recurrent Unit network to not only learn how to predict sequence values but also learn to select the best-performing sequential configuration for each individual feeder. The proposed method was tested on an actual urban distribution system in West Canada. It was compared with traditional methods and our previous sequence prediction method. It demonstrates the best forecasting performance as well as the possibility of using sequence prediction models for multi-year component-level load forecast., Comment: 22 pages, 9 figures

In this paper, we investigate the performance of several sequence prediction techniques on the prediction of future events of human behavior in a smart home, as well as the timestamps of those next events. Prediction techniques in smart home environments have several use cases, such as the real-time identification of abnormal behavior, identifying coachable moments for e-coaching, and a plethora of applications in the area of home automation. We give an overview of several sequence prediction techniques, including techniques that originate from the areas of data mining, process mining, and data compression, and we evaluate the predictive accuracy of those techniques on a collection of publicly available real-life datasets from the smart home environments domain. This contrast our work with existing work on prediction in smart homes, which often evaluate their techniques on a single smart home instead of a larger collection of logs. We found that LSTM neural networks outperform the other prediction methods on the task of predicting the next activity as well as on the task of predicting the timestamp of the next event. However, surprisingly, we found that it is very dependent on the dataset which technique works best for the task of predicting a window of multiple next activities.

Deep recurrent neural networks (DRNNs) have recently demonstrated strong performance in sequential data analysis, such as natural language processing. These capabilities make them a promising tool for inferential analysis of sequentially structured bioinformatics data as well. Here, we assessed the ability of Long Short-Term Memory (LSTM) networks, a class of DRNNs, to predict properties of proteins based on their primary structures. The proposed architecture is trained and tested on two different datasets to predict whether a given sequence falls into a certain class or not. The first dataset, directly imported from Uniprot, was used to train the network on whether a given protein contained or did not contain a conserved sequence (homeodomain), and the second dataset, derived by literature mining, was used to train a network on whether a given protein binds or doesn't bind to Artemisinin, a drug typically used to treat malaria. In each case, the model was able to differentiate between the two different classes of sequences it was given with high accuracy, illustrating successful learning and generalization. Upon completion of training, an ROC curve was created using the homeodomain and artemisinin validation datasets. The AUC of these datasets was 0.80 and 0.87 respectively, further indicating the models' effectiveness. Furthermore, using these trained models, it was possible to derive a protocol for sequence detection of homeodomain and binding motif, which are well-documented in literature, and a known Artemisinin binding site, respectively [1-3]. Along with these contributions, we developed a python API to directly connect to Uniprot data sourcing, train deep neural networks on this primary sequence data using TensorFlow, and uniquely visualize the results of this analysis. Such an approach has the potential to drastically increase accuracy and reduce computational time and, current major limitations in informatics, from inquiry to discovery in protein function research

Recommendation systems that automatically generate personalized music playlists for users have attracted tremendous attention in recent years. Nowadays, most music recommendation systems rely on item-based or user-based collaborative filtering or content-based approaches. In this paper, we propose a novel mixture hidden Markov model (HMM) for music play sequence prediction. We compare the mixture model with state-of-the-art methods and evaluate the predictions quantitatively and qualitatively on a large-scale real-world dataset in a Kaggle competition. Results show that our model significantly outperforms traditional methods as well as other competitors. We conclude by envisioning a next-generation music recommendation system that integrates our model with recent advances in deep learning, computer vision, and speech techniques, and has promising potential in both academia and industry., Accepted to the 4th International Conference on Artificial Intelligence and Applications (AI 2018)

Particulate matter with aerodynamic diameter less than 2.5 µm (PM2.5) in indoor public spaces such as subway stations, has represented a major public health concern; however, forecasting future sequences of quantitative health risk is an effective method for protecting commuters’ health, and an important tool for developing early warning systems. Despite the existence of several predicting methods, some tend to fail to forecast long-term dependencies in an effective way. This paper aims to implement a multiple sequences prediction of a comprehensive indoor air quality index (CIAI) traced by indoor PM2.5, utilizing different structures of recurrent neural networks (RNN). A standard RNN (SRNN), long short-term memory (LSTM) and a gated recurrent unit (GRU) structures were implemented due to their capability of managing sequential, and time-dependent data. Hourly indoor PM2.5 concentration data collected in the D-subway station, South Korea, were utilized for the validation of the proposed method. For the selection of the most suitable predictive model (i.e. SRNN, LSTM, GRU), a point-by-point prediction on the PM2.5 was conducted, demonstrating that the GRU structure outperforms the other RNN structures (RMSE = 21.04 µg/m3, MAPE = 32.92%, R2 = 0.65). Then, this model is utilized to sequentially predict the concentration and quantify the health risk (i.e. CIAI) at different time lags. For a 6-h time lag, the proposed model exhibited the best performance metric (RMSE = 29.73 µg/m3, MAPE = 29.52%). Additionally, for the rest of the time lags including 12, 18 and 24 h, achieved an acceptable performance (MAPE = 29–37%).

This chapter will discuss the concepts of recurrent neural networks (RNNs) and their modified version, long short-term memory (LSTM). LSTM is mainly used for sequence prediction. You will learn about the varieties of sequence prediction and then learn how to do time-series forecasting with the help of the LSTM model.

In order to alleviate data sparsity and overfitting problems in maximum likelihood estimation (MLE) for sequence prediction tasks, we propose the Generative Bridging Network (GBN), in which a novel bridge module is introduced to assist the training of the sequence prediction model (the generator network). Unlike MLE directly maximizing the conditional likelihood, the bridge extends the point-wise ground truth to a bridge distribution conditioned on it, and the generator is optimized to minimize their KL-divergence. Three different GBNs, namely uniform GBN, language-model GBN and coaching GBN, are proposed to penalize confidence, enhance language smoothness and relieve learning burden. Experiments conducted on two recognized sequence prediction tasks (machine translation and abstractive text summarization) show that our proposed GBNs can yield significant improvements over strong baselines. Furthermore, by analyzing samples drawn from different bridges, expected influences on the generator are verified.

Reinforcement learning (RL), which has been successfully applied to sequence prediction, introduces \textit{reward} as sequence-level supervision signal to evaluate the quality of a generated sequence. Existing RL approaches use the ground-truth sequence to define reward, which limits the application of RL techniques to labeled data. Since labeled data is usually scarce and/or costly to collect, it is desirable to leverage large-scale unlabeled data. In this paper, we extend existing RL methods for sequence prediction to exploit unlabeled data. We propose to learn the reward function from labeled data and use the predicted reward as \textit{pseudo reward} for unlabeled data so that we can learn from unlabeled data using the pseudo reward. To get good pseudo reward on unlabeled data, we propose a RNN-based reward network with attention mechanism, trained with purposely biased data distribution. Experiments show that the pseudo reward can provide good supervision and guide the learning process on unlabeled data. We observe significant improvements on both neural machine translation and text summarization.

Recurrent neural networks have recently been successfully applied to the session-based recommendation problem, and is part of a growing interest for collaborative filtering based on sequence prediction. This new approach to recommendations reveals an aspect that was previously overlooked: the difference between short-term and long-term recommendations. In this work we characterize the full short-term/long-term profile of many collaborative filtering methods, and we show how recurrent neural networks can be steered towards better short or long-term predictions. We also show that RNNs are not only adapted to session-based collaborative filtering, but are perfectly suited for collaborative filtering on dense datasets where it outperforms traditional item recommendation algorithms.

Predictive business process monitoring methods exploit logs of completed cases of a process in order to make predictions about running cases thereof. Existing methods in this space are tailor-made for specific prediction tasks. Moreover, their relative accuracy is highly sensitive to the dataset at hand, thus requiring users to engage in trial-and-error and tuning when applying them in a specific setting. This paper investigates Long Short-Term Memory (LSTM) neural networks as an approach to build consistently accurate models for a wide range of predictive process monitoring tasks. First, we show that LSTMs outperform existing techniques to predict the next event of a running case and its timestamp. Next, we show how to use models for predicting the next task in order to predict the full continuation of a running case. Finally, we apply the same approach to predict the remaining time, and show that this approach outperforms existing tailor-made methods., Accepted at the International Conference on Advanced Information Systems Engineering (CAiSE) 2017

This paper focuses on the problem of distributed sequence prediction in a network of sparsely interconnected agents, where agents collaborate to achieve provably reasonable predictive performance. An expert assisted online learning algorithm in a distributed setup of the consensus+innovations form is proposed, in which the agents update their weights for the experts' predictions by simultaneously processing the latest network losses (innovations) and the cumulative losses obtained from neighboring agents (consensus). This paper characterizes the regret of the agents' prediction in lieu of the proposed distributed online learning algorithm and establishes the sub-linear regret of the agents' predictions with respect to the best forecasting expert.

We consider the setting of sequential prediction of arbitrary sequences based on specialized experts. We first provide a review of the relevant literature and present two theoretical contributions: a general analysis of the specialist aggregation rule of Freund et al. (Proceedings of the Twenty-Ninth Annual ACM Symposium on the Theory of Computing (STOC), pp. 334---343, 1997) and an adaptation of fixed-share rules of Herbster and Warmuth (Mach. Learn. 32:151---178, 1998) in this setting. We then apply these rules to the sequential short-term (one-day-ahead) forecasting of electricity consumption; to do so, we consider two data sets, a Slovakian one and a French one, respectively concerned with hourly and half-hourly predictions. We follow a general methodology to perform the stated empirical studies and detail in particular tuning issues of the learning parameters. The introduced aggregation rules demonstrate an improved accuracy on the data sets at hand; the improvements lie in a reduced mean squared error but also in a more robust behavior with respect to large occasional errors.

Although most quantitative financial data are analyzed using traditional statistical, artificial intelligence or data mining techniques, the abundance of online electronic financial news articles has opened up new possibilities for intelligent systems that can extract and organize relevant knowledge automatically in a usable format. Most information extraction systems require a hand-built dictionary of templates and thus need continual modification to accommodate new patterns that are observed in the text. In this research, we propose a novel text-based decision support system (DSS) that (i) extracts event sequences from shallow text patterns, and (ii) predicts the likelihood of the occurrence of events using a classifier-based inference engine. The prediction relies on two major, but complementary, feature sets: adjacent events and a set of information-theoretic functions. In contrast to other approaches, the proposed text-based DSS gives explanatory hypotheses about its predictions from a coalition of intimations learned from the inference engine, while preserving robustness and without indulging in formalism. We investigate more than 2000 financial reports with 28,000 sentences. Experiments show that the prediction accuracy of our model outperforms similar statistical models by 7% for the seen data while significantly improving the prediction accuracy for the unseen data. Further comparisons substantiate the experimental findings.

Recently, attention has been drawn to the local modeling techniques of a new idea called “Just-In-Time (JIT) modeling”. To apply “JIT modeling” to a large amount of database online, “Large-scale database-based Online Modeling (LOM)” has been proposed. LOM is a technique that makes the retrieval of neighboring data more efficient by using both “stepwise selection” and quantization. In order to predict the long-term state of the plant without using future data of manipulated variables, an Extended Sequential Prediction method of LOM (ESP-LOM) has been proposed. In this paper, the LOM and the ESP-LOM are introduced.