Artificial intelligence has acted as an essential driver of emerging technologies by employing many sophisticated Machine Learning (ML) models, while lack of model transparency and results explanation limits its effectiveness in real decision-making. The eXplainable AI (XAI) has bridged this gap by providing the explanation of outcomes made by these complex ML model. In this paper, we classify the functioning of an air handling unit (AHU) using the neural network and utilise contextual importance and contextual utility (CIU) as an XAI module for explaining outcome of the neural Network. Here, we prove that CIU (XAI module) can generate transparent and human-understandable explanations, which the end-user can therefore utilize for making decisions proving the overall applicability of the method in a novel use-case. Visual and textual explanations for the causes of an individual prediction have been derived from the CIU that are numeric values calculated from the machine learning module results. We also have provided contrasting explanations against some causes that were not involved in the decision. We provide both in our proposed approach.

Plant diseases are one of the biggest challenges faced by the agricultural sector due to the damage and economic losses in crops. Despite the importance, crop disease diagnosis is challenging because of the limited-resources farmers have. Subsequently, the early diagnosis of plant diseases results in considerable improvement in product quality. The aim of the proposed work is to design an ML-powered mobile-based system to diagnose and provide an explanation based remedy for the diseases in grape leaves using image processing and explainable artificial intelligence. The proposed system will employ the computer vision empowered with Machine Learning (ML) for plant disease recognition and explains the predictions while providing remedy for it. The developed system uses Convolutional Neural networks (CNN) as an underlying machine/deep learning engine for classifying the top disease categories and Contextual Importance and Utility (CIU) for localizing the disease areas based on prediction. The user interface is developed as an IOS mobile app, allowing farmers to capture a photo of the infected grape leaves. The system has been evaluated using various performance metrics such as classification accuracy and processing time by comparing with different state-of-the-art algorithms. The proposed system is highly compatible with the Apple ecosystem by developing IOS app with high prediction and response time. The proposed system will act as a prototype for the plant disease detector robotic system.

Explaining the result of machine learning models is an active research topic in Artificial Intelligence (AI) domain with an objective to provide mechanisms to understand and interpret the results of the underlying black-box model in a human-understandable form.With this objective, several eXplainable Artificial Intelligence (XAI) methods have been designed and developed based on varied fundamental principles.Some methods such as Local interpretable model agnostic explanations (LIME), SHAP (SHapley Additive exPlanations) are based on the surrogate model while others such as Contextual Importance and Utility (CIU) do not create or rely on the surrogate model to generate its explanation.Despite the difference in underlying principles, these methods use different sampling techniques such as uniform sampling, weighted sampling for generating explanations.CIU, which emphasizes a context-aware decision explanation, employs a uniform sampling method for the generation of representative samples.In this research, we target uniform sampling methods which generate representative samples that do not guarantee to be representative in the presence of strong non-linearities or exceptional input feature value combinations.The objective of this research is to develop a sampling method that addresses these concerns.To address this need, a new adaptive weighted sampling method has been proposed.In order to verify its efficacy in generating explanations, the proposed method has been integrated with CIU, and tested by deploying the special test case.

Different explainable AI (XAI) methods are based on different notions of 'ground truth'. In order to trust explanations of AI systems, the ground truth has to provide fidelity towards the actual behaviour of the AI system. An explanation that has poor fidelity towards the AI system's actual behaviour can not be trusted no matter how convincing the explanations appear to be for the users. The Contextual Importance and Utility (CIU) method differs from currently popular outcome explanation methods such as Local Interpretable Model-agnostic Explanations (LIME) and Shapley values in several ways. Notably, CIU does not build any intermediate interpretable model like LIME, and it does not make any assumption regarding linearity or additivity of the feature importance. CIU also introduces the value utility notion and a definition of feature importance that is different from LIME and Shapley values. We argue that LIME and Shapley values actually estimate 'influence' (rather than 'importance'), which combines importance and utility. The paper compares the three methods in terms of validity of their ground truth assumption and fidelity towards the underlying model through a series of benchmark tasks. The results confirm that LIME results tend not to be coherent nor stable. CIU and Shapley values give rather similar results when limiting explanations to 'influence'. However, by separating 'importance' and 'utility' elements, CIU can provide more expressive and flexible explanations than LIME and Shapley values.

Online advertisement has become a major commercial campaign in social networks. Many big companies have invested massive resources for collecting data about the users and their web surfing habits. Utilising these data, the advertisement companies can get valuable insights about the users and their interests. The gathered information can improve the effectiveness of advertisement campaigns by identifying potential customers of a product/service or by identifying purchase patterns. A successful advertisement campaign depends on the company's ability to fully leverage these data assets. As the artificial intelligence flourish with the machine learning models which were offered as a solution for such a problem depending on dataset availability and computation power but the resulting systems suffer from a loss of transparency and interpretability, especially for end-users.In order to overcome the aforementioned problem of explainability of the models, we propose an explainable and interpretable approach to solve this problem. In the first stage, machine learning model will be used to develop a predictive model that is capable of predicting potential customers who are likely to click the advertisement of a particular product/services. This approach is tested on the public advertising dataset. In the second stage, the predictive model is further utilised by local surrogate model initially using Local Interpretable Model-agnostic Explanations (LIME) to locally approximating the model around a given prediction and then with global interpretable explanations by considering whole machine learning model at once. Finally, Contextual Importance and Utility (CIU) is used for global explanations to generate the explanations and interpretation of the prediction based on the contributing features of the dataset.

We are entering a new age of AI applications where machine learning is the core technology but machine learning models are generally non-intuitive, opaque and usually complicated for people to understand. The current AI applications inability to explain is decisions and actions to end users have limited its effectiveness. The explainable AI will enable the users to understand, accordingly trust and effectively manage the decisions made by machine learning models. The heat recycler’s fault detection in Air Handling Unit (AHU) has been explained with explainable artificial intelligence since the fault detection is particularly burdensome because the reason for its failure is mostly unknown and unique. The key requirement of such systems is the early diagnosis of such faults for its economic and functional efficiency. The machine learning models, Support Vector Machine and Neural Networks have been used for the diagnosis of the fault and explainable artificial intelligence has been used to explain the models’ behaviour.

Information Systems (ISs) are fundamental to streamline operations and support processes of any modern enterprise. Being able to perform analytics over the data managed in various enterprise ISs is becoming increasingly important for organisational growth. Extract, Transform, and Load (ETL) are the necessary pre-processing steps of any data mining activity. Due to the complexity of modern IS, extracting data is becoming increasingly complicated and time-consuming. In order to ease the process, this paper proposes a methodology and a pilot implementation, that aims to simplify data extraction process by leveraging the end-users' knowledge and understanding of the specific IS. This paper first provides a brief introduction and the current state of the art regarding existing ETL process and techniques. Then, it explains in details the proposed methodology. Finally, test results of typical data-extraction tasks from four commercial ISs are reported.

This paper proposes an explainable machine learning tool that can potentially be used for decision support in medical image analysis scenarios. For a decision-support system it is important to be able to reverse-engineer the impact of features on the final decision outcome. In the medical domain, such functionality is typically required to allow applying machine learning to clinical decision making. In this paper, we present initial experiments that have been performed on in-vivo gastral images obtained from capsule endoscopy. Quantitative analysis has been performed to evaluate the utility of the proposed method. Convolutional neural networks have been used for training the validating of the image data set to provide the bleeding classifications. The visual explanations have been provided in the images to help health professionals trust the black box predictions. While the paper focuses on the in-vivo gastral image use case, most findings are generalizable.

Industrial Internet of things is becoming a boon to Original Equipment Manufacturers (OEMs) offering after-sales services such as condition-based maintenance and extended warranty for their products. These companies leverage novel digital information infrastructures to improve daily industrial activities, including data collection, remote monitoring and advanced condition-based maintenance services. The emergence of digital infrastructure and new business prospects via servi-tization and quality services encourage companies to collect vast amounts of data that have been generated in different stages of product lifecycles. Despite of the potential benefits, companies are unable to fully harness the opportunities presented by digital information infrastructure because there exist several platforms with variations in technologies and standards resulting in interoperability challenges. This becomes particularly critical when a company sells its products to several clients with different technologies. To overcome such challenges, we investigate the Open Messaging Interface (O-MI) and Open Data Format (O-DF), flexible messaging and data exchange standards that enable seamless integration of different systems. These standards enable interoperability and support time-centric, event-centric, and rate-centric modes of data exchange.

Many domains are trying to integrate with the Internet of Things (IoT) ecosystem, such as public administrations starting smart city initiatives all over the world. Cities are becoming smart in many ways: smart mobility, smart buildings, smart environment and so on. However, the problem of non-interoperability in the IoT hinders the seamless communication between all kinds of IoT devices. Different domain specific IoT applications use different interoperability standards. These standards are usually not interoperable with each other. IoT applications and ecosystems therefore tend to use a vertical communication model that does not allow data sharing horizontally across different IoT ecosystems. In 2014, The Open Group published two domain-independent IoT messaging standards, O-MI and O-DF, aiming to solve the interoperability problem. In this article we describe the practical use of O-MI/O-DF standards for reaching interoperability in a mobile application for the smart city context, in particular for the Smart Mobility domain, electric vehicle (EV) charging case study. The proof-of-concept of the smart EV charging ecosystem with mobile application user interface was developed as a part of an EU (Horizon 2020) Project bIoTope.

IoT systems may provide information from different sensors that may reveal potentially confidential data, such as a person’s presence or not. The primary question to address is how we can identify the sensors and other devices in a reliable way before receiving data from them and using or sharing it. In other words, we need to verify the identity of sensors and devices. A malicious device could claim that it is the legitimate sensor and trigger security problems. For instance, it might send false data about the environment, harmfully affecting the outputs and behavior of the system. For this purpose, using only primary identity values such as IP address, MAC address, and even the public-key cryptography key pair is not enough since IPs can be dynamic, MACs can be spoofed, and cryptography key pairs can be stolen. Therefore, the server requires supplementary security considerations such as contextual features to verify the device identity. This paper presents a measurement-based method to detect and alert false data reports during the reception process by means of sensor behavior. As a proof of concept, we develop a classification-based methodology for device identification, which can be implemented in a real IoT scenario.

Maintenance is a complicated task that encompasses various activities including fault detection, fault diagnosis, and fault reparation. The advancement of Computer Aided Engineering (CAE) has increased challenges in maintenance as modern assets have became complex mixes of systems and sub systems with complex interaction. Among maintenance activities, fault diagnosis is particularly cumbersome as the reason of failures on the system is often neither obvious in terms of their source nor unique. Early detection and diagnosis of such faults is turning to one of the key requirements for economical and functional asset efficiency. Several methods have been investigated to detect machine faults for a number of years that are relevant for many application domains. In this paper, we present the process history-based method adopting nominal efficiency of Air Handling Unit (AHU) to detect heat recovery failure using Principle Component Analysis (PCA) in combination of the logistic regression method.

The number of Internet of Things (IoT) vendors is rapidly growing, providing solutions for all levels of the IoT stack. Despite the universal agreement on the need for a standardized technology stack, following the model of the world-wide-web, a large number of industry-driven domain specific standards hinder the development of a single IoT ecosystem. An attempt to solve this challenge is the introduction of O-MI (Open Messaging Interface) and O-DF (Open Data Format), two domain independent standards published by Open Group. Despite their good compatibility, they define no specific security model. This paper takes the first step of defining a security model for these standards by proposing suitable access control and authentication mechanisms that can regulate the rights of different principles and operations defined in these standards. First, a brief introduction is provided of the O-MI and O-DF standards, including a comparison with existing standards. Second, the envisioned security model is presented, together with the implementation details of the plug-in module developed for the O-MI and O-DF reference implementation.

In today's competitive and fluctuating market, original equipment manufacturers (OEMs) must be able to offer aftersales services along with their products, such as condition based maintenance, extended warranty services etc. Condition based maintenance requires detailed understanding about products' operational behaviour, to detect problems before they occur, and react accordingly. Typically, Condition based maintenance consists of data collection, data analysis, and maintenance decision stages. Within this context, data quality is one of the key drivers in the knowledge acquisition process since poor data quality impacts the downstream maintenance processes, and reciprocally, high data quality will foster good decision making. The prospect of new business opportunities and better services to customers encourages companies to collect large amounts of data that have been generated in different stages of product lifecycle. Despite of availability of data, as well as advanced statistical and analytical tools, companies are still struggling to provide effective service by reducing maintenance cost and improving uptime. This paper highlights data related pitfalls that hinder organisations to improve maintenance services. These pitfalls are based on case studies of two globally operating Finnish manufacturing companies where maintenance is one of the major streams of income.

Information Systems (ISs) are fundamental to streamline operations and support processes of any modern enterprise. Being able to perform analytics over the data managed in various enterprise ISs is becoming increasingly important for organisational growth. Extract, Transform, and Load (ETL) are the necessary pre-processing steps of any data mining activity. Due to the complexity of modern IS, extracting data is becoming increasingly complicated and time-consuming. In order to ease the process, this paper proposes a methodology and a pilot implementation, that aims to simplify data extraction process by leveraging the end-users' knowledge and understanding of the specific IS.

Service business renewal to support global fleet operations is one of the critical competitiveness factors of technology industries. This paper describes preliminary results of defining an IoT-based framework for fleet management systems (FMS). The main contribution is to provide a systematic outline to the ingredients of the framework and discussing their impact in an industrial setting. The overall aim of the research is to define a technical framework for fleet management solutions and to define the necessary set of functionalities that the framework should support as a base for FMS systems. The framework will be implemented and validated in collaboration with industrial partners.

Businesses are increasingly using their enterprise data for strategic decision-making activities. In fact, information, derived from data, has become one of the most important tools for businesses to gain competitive edge. Data quality assessment has become a hot topic in numerous sectors and considerable research has been carried out in this respect, although most of the existing frameworks often need to be adapted with respect to the use case needs and features. Within this context, this paper develops a methodology for assessing the quality of enterprises' daily maintenance reporting, relying both on an existing data quality framework and on a Multi-Criteria Decision Making (MCDM) technique. Our methodology is applied in cooperation with a Finnish multinational company in order to evaluate and rank different company sites/office branches (carrying out maintenance activities) according to the quality of their data reporting. Based on this evaluation, the industrial partner wants to establish new action plans for enhanced reporting practices.

The Industrial Internet promises to radically change and improve many industry's daily business activities, from simple data collection and processing to context-driven, intelligent and pro-active support of workers' everyday tasks and life. The present paper first provides insight into a typical industrial internet application architecture, then it highlights one fundamental arising contradiction: "Who owns the data is often not capable of analyzing it". This statement is explained by imaging a visionary data supply chain that would realize some of the Industrial Internet promises. To concretely implement such a system, recent standards published by The Open Group are presented, where we highlight the characteristics that make them suitable for Industrial Internet applications. Finally, we discuss comparable solutions and concludes with new business use cases.