Intelligent interface for engineering design

As the complexity of intelligent public transportation systems continues to increase and the degree of interconnection of equipment at all levels gradually deepens, building a digital twin system will generate a large amount of multi-source heterogeneous data. Effective processing and in-depth mining of data can provide real-time and intelligent decision-making optimization for the construction of digital twin systems, thereby efficiently handling complex emergencies. This paper conducts a systematic review on the processing methods and technologies of multi-source heterogeneous twin data in traffic dynamic scenarios. First, the content and classification of multi-source heterogeneous twin data in the operation process of intelligent bus systems are clarified; secondly, the multi-source The data processing methods and technologies applied in each stage of data collection, data storage, data fusion and data analysis in heterogeneous twin data processing are analyzed, and the advantages, disadvantages and applications of various methods and technologies are analyzed; finally, corresponding to the multiple This paper summarizes the methods and technologies for processing multi-source heterogeneous twin data, and points out the challenges and development trends faced by the current multi-source heterogeneous twin data processing methods and technologies.

Automated user modeling for intelligent interface

The computational processing methods appeared to overcome the drawbacks of the symbolic processing method. Recently the serious limitations of the computational processing methods have been found through the development of the highly intelligent user interface systems. The simple solutions to the limitations are the return to the symbolism or the use of the hierarchical system architecture. However it is indispensable deeply to combine the symbolic and computational processing in order to realize the highly intelligent system. The leading attempts toward the combination of computational and symbolic processing are overviewed. The level of the combination is not deep, high and wide enough. Our approach based on the emergent recurrent neural network and the language processing toward the deep fusion of symbolic and computational processing is proposed. In this new architecture, the symbolic processing and the computational processing should affect each other at anytime, at anywhere and at any abstract level.

A domain expert can process heterogeneous data to make meaningful interpretations or predictions from the data. For example, by looking at research papers and patent records, an expert can determine the maturity of an emerging technology and predict the geographic location(s) and time (e.g., in a certain year) where and when the technology will be a success. However, this is an expert- and manual-intensive task. This paper presents an end-to-end system that integrates heterogeneous data sources into a knowledge graph in the RDF (Resource Description Framework) format using an ontology. Then the user can easily query the knowledge graph to prepare the required data for different types of predictive analysis tools. We show a case study of predicting the (geographic) center(s) of fuel cell technologies using data collected from public sources to demonstrate the feasibility of our system. The system extracts, cleanses, and augments data from public sources including research papers and patent records. Next, the system uses an ontology-based data integration method to generate knowledge graphs in the RDF format to enable users to switch quickly between machine learning models for predictive analytic tasks. We tested the system using the Support Vector Machine and Multiple Hidden Markov Models and achieved 66.7% and 83.3% accuracy on the city and year levels of spatial and temporal resolutions, respectively.

Automatic media conversion schemes from speech to a facial image and a construction of a real-time image synthesis system are presented. The purpose of this research is to realize an intelligent human-machine interface or intelligent communication system with synthesized human face images. A human face image is reconstructed on the display of a terminal using a 3-D surface model and texture mapping technique. Facial motion images are synthesized by transformation of the 3-D model. In the motion driving method, based on vector quantization and the neural network, the synthesized head image can appear to speak some given words and phrases naturally, in synchronization with voice signals from a speaker. >

With the rapid development of information technology, large amounts of multi-source data are constantly being generated in medical field. The automatic visualization system based on them has gained a lot of attention, since the intuitive data presentation can help even non-professional users effectively get the information hidden behind the separate data obtained from different scenarios and make better decisions. In this paper, based on the Data Lake architecture, we improve the performance of an existing novel data visualization recommendation system and resolve three challenges about the processing of multi-source and heterogeneous data. First, we build the framework based on Data Lake to store multi-source and heterogeneous data. Second, we optimize the data manipulation module in the visualization system based on the distributed processing power of Data Lake to get potentially interesting visualization candidates in a short time. Third, we efficiently run exploratory queries on large datasets based on the calculation capability of Data Lake to meet the actual needs of users. According to the experiment results, our system demonstrates a remarkable acceleration effect on the task of automatic visualization of big multi-source medical data.

With the continuous development of smart cities, intelligent transportation systems (ITSs) have ushered in many breakthroughs and upgrades. As a solid foundation for an ITS, traffic flow prediction effectively helps the city to better manage intricate traffic flow. However, existing traffic flow prediction methods such as temporal graph convolutional networks(T-GCNs) ignore the dissimilarities between lanes. Thus, they cannot provide more specific information regarding predictions such as dynamic changes in traffic flow direction and deeper lane relationships. With the upgrading of intersection sensors, more and more intersection lanes are equipped with intersection sensors to detect vehicle information all day long. These spatio-temporal data help researchers refine the focus of traffic prediction research down to the lane level. More accurate and detailed data mean that it is more difficult to mine the spatio-temporal correlations between data, and modeling heterogeneous data becomes more challenging. In order to deal with these problems, we propose a heterogeneous graph convolution model based on dynamic graph generation. The model consists of three components. The internal graph convolution network captures the real-time spatial dependency between lanes in terms of generated dynamic graphs. The external heterogeneous data fusion network comprehensively considers other parameters such as lane speed, lane occupancy, and weather conditions. The codec neural network utilizes a temporal attention mechanism to capture the deep temporal dependency. We test the performance of this model based on two real-world datasets, and extensive comparative experiments indicate that the proposed heterogeneous graph convolution model can improve the prediction accuracy.

A method for fraud detection when installing mobile applications was proposed. The developed method, in contrast to existing ones, uses all available data regardless of their types, dimensions, and discrepancies and converts such data into homogeneous coefficients based on the proposed scaling method. This approach allows one to improve accuracy of task solution and build an open to expansion knowledge base with characteristics of fraudsters and rules of detecting fraudulent users. A system of scales for converting heterogeneous data into homogeneous coefficients has been developed which has enabled construction of a mathematical model of the scaling process. The algorithm of scaling heterogeneous data sets based on the proposed scales and the mathematical model of the process of scaling large arrays of heterogeneous data has been developed which has made it possible to reduce the whole data set to two homogeneous groups. The algorithms of processing the resulting groups of homogeneous data and detection of fraudulent users were offered. The developed algorithms using coefficients of similarity between user characteristics form fingerprints of fraudsters, determine characteristics and dependences of fraudsters which allows one to increase efficiency and speed of the process of fraudster detection. A scheme of the fraud detection process which was used in the intelligent system of automatic detection of fraudsters for carrying out of experimental studies was proposed. According to the results of experimental study, accuracy of fraudster detection was 99.14 % for a given representative sample. The results of experimental studies have shown effectiveness of automatic detection of fraudsters and the possibility of expanding formats and characteristics of fraudsters based on intelligent analysis and knowledge bases.

Successful adaptation to climate change requires resilient, reliable, and efficient energy systems. To unlock energy efficiency potentials in buildings, an intelligent, user-centered approach is vital. However, this requires handling diverse data on the energy system. Therefore, technologies for harmonizing, storing, and visualizing data, as well as managing physical devices and users are needed. This work assesses existing and required Information and Communication Technologies (ICT) for intelligent building energy system operation. We propose an intermediate architecture based on Internet of Things (IoT) core principles and feature insights from its implementation within the Living Lab Energy Campus (LLEC) at Forschungszentrum Jülich. We present an approach for integrating existing ICT components, such as building energy metering and central Heating, Ventilation and Air Conditioning (HVAC) management, and propose a comprehensive data collection and distribution infrastructure. We establish IoT-enabled applications for energy system monitoring, user engagement, advanced building operation, and device identification and management. We evaluate our ICT setup through functional and performance assessments. We find that heterogeneous data can be reliably collected, distributed, and managed using standardized interfaces, state-of-the-art databases, and cutting-edge software components. For the buildings operated through the ICT infrastructure, data transmission availability is above 98.90 %, mean time to repair (MTTR) is less than 2.68 h, and mean time between failures (MTBF) is in the range of 242.67 h to 1092.00 h, evaluated over a period of three months. Our approach promotes the early real-world adoption of intelligent building control prototypes and their sustainable development. We demonstrate the proposed ICT setup through an experimental study that applies a cloud-based Model Predictive Controller (MPC) to a real building space. Our results provide a comprehensive discussion of the required ICT setup for intelligent building energy system control in real-world environments, and highlight important design strategies that reduce the conceptual overhead and facilitate implementation in similar projects.

Subject. In the context of the digital transformation of the construction industry, the key barrier to increasing economic efficiency is not the lack of data, but the inability of existing management systems to integrate and analyze heterogeneous information flows to build reliable forecasts. The separation of static information models (BIM) and dynamic equipment telemetry (IoT) generates an asymmetry of management decisions, leading to systematic budget overruns and delays in the implementation of investment and construction projects. Objectives. Development of an economic and mathematical methodology for predictive management of construction processes, which minimizes operational risks and increases the investment attractiveness of projects by integrating deep machine learning methods into the production planning contour. Methods. The research is based on the synthesis of methods of systems analysis, theory of economic risks and technologies of intellectual data processing. Using Monte Carlo stochastic modeling algorithms, a unique dataset has been generated that combines the attribute characteristics of BIM models (IFC classes, material volumes), telemetry flows of construction equipment, and external factors (weather conditions, logistical delays). Recurrent neural networks of the LSTM architecture with mathematical formalization of the optimization process of the objective function of economic losses are used to identify hidden nonlinear dependencies. Results. A methodology for integrating heterogeneous data has been developed and verified, including strict rules for time synchronization (Δt = 1 hour) and semantic binding via globalId identifiers. A comparative analysis of four architectures (linear regression, Random Forest, XGBoost, LSTM) on a test sample demonstrated a statistically significant superiority of recurrent networks: the average absolute prediction error was reduced to 1.04 days (MAPE = 4.1%) with a coefficient of determination R2 = 0.94, which is 3 times more accurate than traditional methods and 38% more accurate than ensemble methods. algorithms of gradient boosting. Conclusions. The introduction of intelligent predictive management systems integrating BIM and IoT ensures the transition from reactive response to proactive risk management, creating an economic effect by reducing unproductive downtime, optimizing resource use and minimizing penalties for failure to comply with contractual obligations. The proposed methodology lays the foundation for the creation of self-learning digital counterparts of construction sites and can be scaled to enterprises of the investment and construction complex operating in a highly volatile environment.

AI’s 10 to Watch, 2022

Aimed at the characteristics of huge systems, complicated distributed structures and difficult fault diagnosis of large equipments, an extensible system of distributed online monitoring and multi-level intelligent fault diagnosis based on hybrid wireless network techniques was presented and designed. The entire system includes four hierarchy's subsystems. The first subsystem is monitoring sensors and Zigbee network. It is responsible to gain various data of equipments in a non-intrusive mode. The second one is monitoring and transmission network. It consists of embedded system ARM920T and relative multiple single-chips, and is responsible to receive and integrate all information of the previous subsystem. The third one is online and offline monitoring subsystem based on wireless protocol 802.11b. The final one is multi-level intelligent diagnosis and management system. The results of practical operation show that, because of the using of hybrid wireless network, embedded system and multilevel intelligent diagnosis, the system may be flexibly deployed, have powerful capability of monitoring, diagnosis and management, and can implement expeditious and accurate fault positioning and diagnosis for large equipment system.

It is our great pleasure to welcome you to the 2012 ACM International Conference on Intelligent User Interfaces -- IUI'12 and to Lisboa, Portugal. Starting in 1993, IUI is now on its 17th edition and has established itself has the premier venue for reporting outstanding research and development on intelligent user interface. The IUI series continues to be the principal forum for the meeting of the Human-Computer Interaction (HCI) and the Artificial Intelligence (AI) communities. This gives rise to unique, creative, interdisciplinary contributions. This year's meeting showcases those topics, with works covering innovations in mobile interfaces, collaboration technology, affective interfaces, multimodal interfaces, pen interfaces, haptic and gesture interfaces, multi-touch interfaces, speech interfaces, gaze-based interfaces, tabletop interfaces, health applications, entertainment applications, adaptive narratives and theater, sports and in-vehicle applications, geographic applications, social media interfaces, educational interfaces, sketch recognition, human-robot interfaces, personalization and assistive technologies, end-user programming, ubiquitous and smart environments, locationaware interfaces, recommender interfaces, persuasive interfaces, web-based interfaces and agentbased interfaces. Building on previous years' success, the IUI 2012 call for papers attracted 134 full paper submissions and 78 short paper submissions. To ensure the highest possible quality we assembled a team of 48 senior program committee members and 367 reviewers. This guaranteed that every submission had at least three reviews plus one meta-review. This was followed by a rebuttal phase, keeping the procedure started two years ago. We believe that this process has guaranteed the highest quality possible for this year technical program and we would like to thank all the senior program committee members and reviewers for their hard work in making sure the most relevant works were selected. Through this very thorough review process we were able to accept 18 full papers, 15 short papers and 16 poster presentations, meaning an acceptance rate of 13% for full papers, 16% for oral presentations and 23% for combined oral and poster presentations. In addition to the full and short papers presentations and poster presentations, IUI will also feature a Demonstrations session for which 18 submissions were accepted from a total of 26 submissions. This year, the IUI programme includes three notable invited speakers sharing their innovative work and experiences with the conference participants: Alex 'Sandy' Pentland from the MIT Human Dynamic Lab, Christopher Bishop from Microsoft Research and Takeo Igarashi from the University of Tokyo. In order to foster the growth of the community, IUI 2012 will feature for the first time in the IUI series a Doctoral Consortium. Its success can be measured by the 12 high quality submissions received. Doctoral Consortium students will also be able to exhibit their work to the main conference audience in the poster session, thus promoting their insertion into the IUI community. We would also like to thank the conference sponsors and supporters that made it possible to provide financial support for Doctoral Consortium attendees and other students participating in the conference. Nine full-day workshops will take place at IUI 2012 covering diverse trending topics in the IUI area: Activity Context Representation: Techniques and Languages; Developing Intelligent User Interfaces for e-Accessibility and e-Inclusion; Context-awareness in Retrieval and Recommendation (CaRR 2012); 3rd Workshop on Semantic Models for Adaptive Interactive Systems (SEMAIS); Scent and Scensibility; User Modeling from Social Media; 2nd Workshop on Interacting with Smart Objects; 2nd Workshop on Location Awareness for Mixed and Dual Reality (LaMDa'12); 1st International Workshop on Ubiquitous Personalization (UP'2012). IUI 2012 continues and extends the cooperation with the ACM Transactions on Interactive Intelligent Systems (TiiS). Besides the streamlined processing of journal submissions based on IUI 2012 papers, IUI 2012 will feature a special session where authors of recently accepted articles for the ACM TiiS will spotlight key contributions of their work. It is our conviction that it will benefit greatly every member of the IUI community. We would like to thank Anthony Jameson and John Riedl, the Editors in Chief of TiiS, for their cooperation.

To survey person centered survival rate in population based screening program by an intelligent clinical decision support system. Colorectal cancer is the most common malignancy and major cause of morbidity and mortality throughout the world. Colorectal cancer is the sixth leading cause of cancer death in Iran. In this survey, we used cosine similarity as data mining technique and intelligent system for estimating survival of at risk groups in the screening plan. In the first step, we determined minimum data set (MDS). MDS was approved by experts and reviewing literatures. In the second step, MDS were coded by python language and matched with cosine similarityformula. Finally, survival rate by percent was illustrated in the user interface of national intelligent system. The national intelligent system was designed in PyCharm environment. Main data elements of intelligent system consist demographic information, age, referral type, risk group, recommendation and survival rate. Minimum data set related to survival comprise of clinical status, past medical history and socio-demographic information. Information of the covered population as a comprehensive database was connected to intelligent system and survival rate estimated for each patient. Mean range of survival of HNPCC patients and FAP patients were respectively 77.7% and 75.1%. Also, the mean range of the survival rate and other calculations have changed with the entry of new patients in the CRC registry by real-time. National intelligent system monitors the entire of risk group and reports survival rates by electronic guidelines and data mining technique and also operates according to the clinical process. This web base software has a critical role in the estimation survival rate in order to health care planning.

This paper aims at the implementation of online e-learning system for spoken Chinese based on the approach of natural language processing. The system is composed of four components: text analysis, prosody estimation, selection of synthesis units, speech generation module. The corpora, dictionary and speech database are generated. More than 4000 monosyllabic speech units of Chinese and several silences with various durations have been recorded as basic synthesis units. Twelve college students with frequent speaking ability evaluated the synthesized speech for the quality of energy and speech intelligibility. The satisfaction results for the testing achieve 8.09 and 7.63, respectively. Based on the comprehensive evaluation, it is obvious that our intelligent system can provide real speech with natural quality and helpful lexical information on Internet.