Despite recent advances in using Decision Support Systems (DSS) for automating or enhancing processes within organizations, their adoption remains low in work contexts. While much research has explored DSS adoption by individual workers, there has been limited focus on the broader multi-stakeholder systems in which they operate. To study this further, this research investigates the adoption of a DSS within a multi-stakeholder environment by using a case study on passenger flow control situated at an international airport.

After the literature research, a context research was performed, consisting of observations, unstructured interviews, and a semi-structured interview study with 11 participants. During this research, several challenges faced by the flow controllers were identified, alongside tensions present between flow controllers, flow guiders, and flow moderators.

By discussing the effects that introducing a DSS might have on these existing challenges and tensions, potential opportunities and adoption barriers were formulated for the adoption of the DSS. The potential adoption barriers were identified on two levels: the integration of the DSS into the multi-stakeholder system and the interaction between the flow controller and the DSS. Regarding the integration into the multi-stakeholder system, the introduction of the DSS might deteriorate interactions between flow controllers, flow moderators, and flow guiders, limiting the exchange of important information and alignment regarding the decision-making process, which is currently valued. Regarding the interaction between decision-makers and the DSS, decision-makers might struggle to integrate subjective insights with DSS recommendations, as subjective information is not considered by the system.

To address these adoption issues, ideation was conducted using storyboarding. Based on the insights gathered, design guidelines were formulated.

The design guidelines highlight that DSS adoption in multi-stakeholder systems is influenced by the dynamics between stakeholders. And proposed to consider this in the design of DSS, by seeking closer involvement and collaboration with the flow moderators and flow guiders during the decision-making process. Also, considering how flow controllers can be supported in combining subjective and contextual insights with recommendations of the DSS is proposed. The guidelines were validated with the intended user group through testing their actionability and understandability (n=6).

Ultimately, this research contributes to the human-computer interaction (HCI) community by formulating design guidelines that address the complexities of DSS adoption within multi-stakeholder systems. It also provides practical insights for organizations, such as Schiphol, by offering a structured approach to integrating DSS into operational workflows while maintaining stakeholder engagement and collaboration.

Keywords: Decision support systems, adoption, aviation, artificial intelligence, barriers and opportunities, design guidelines

Research has shown how people anthropomorphize conversational agents (CA) and unconsciously bring their gender stereotypes into human-agent interaction. For this reason, there has been a long lasted dilemma on whether designers should design CAs that conform to or violate stereotypical expectations. Despite the urgency and importance of navigating through this dilemma, how to better design the gender identity of CA is still an open research question. In this thesis, we describe the problem space of CA identity design and argue that we can calibrate the gender effect by manipulating a metaphor we attach to the CAs. To this end, we approached these research questions from three angles: (1) developing a framework to address the ethical dilemma in CA identity design, (2) evaluating the effect of gender and metaphor in chatbot profiles, and (3) calibrating gender stereotyping through metaphor manipulation.

Visually impaired people should enjoy the same rights to acquire information as people with normal sight. Since visual contents become more and more pervasive in our daily life, image description becomes increasingly important to help visually impaired people to get equal access to the information contained in visual contents. However, how to produce image descriptions in a scalable and reliable way is still an unsolved problem. Therefore, researches on the requirements of image description from the perspective of visually impaired people are essential to approaching this problem. Based on a review of existing study results on this topic, this thesis investigates the possibilities of utilizing interactive image description as an approach to collect visually impaired people’s requirements on image description and the benefits of integrating interactive image description to the current image description production system. The existing one-shot statistical description requires descriptors to evaluate the importance of the image, make choices on what should be described, and organize the content so that necessary information can be effectively conveyed. Through literature review, it is found that the requirements of image description are highly context-dependent and influenced by plentiful factors. Therefore, existing guidelines are usually vague and require the describer to rely on experience and intuition while making a lot of subjective judgments, which increases the threshold for generating high-quality image descriptions. On the other hand, through field research and literature research it is found that VIPs hope to have more control over the presentation of image description (both its presence and content). Early explorations of interactive image description showed the possibility of this affordance. Since users are allowed to decide the description content actively, it is argued that the user's preference for image description can be collected through interactive image description. A design goal is proposed accordingly. A prototype is developed to verify this proposal. Through a comparative experiment, the systems’ function to collect user preferences and gradually improve the content of image description is confirmed. In addition, the qualitative research results also reveal the mental activities when users interacting with image description and the impact of interactive image description in this procedure, which is summarized as an image perception model. It is also argued that structured description and progressive description provide new perspectives to reduce the workload of describing images. A final design was developed as the demonstrator for the research findings and proposals.

Facial emotion recognition from thermal images has gained more attention in recent years. Thermal cameras capture the heat emitted by objects and therefore thermal images are not sensitive to illumination changes. Furthermore, changes in temperature can indicate emotions and it is harder for humans to fake emotions in front of a thermal camera. However, a limitation is that, thermal cameras that capture high-resolution images are expensive, and cheaper thermal cameras often capture images with a low-resolution and/or contaminated with noise and blur. Besides, low-resolution thermal images can also arise when images are captured from a far distance or from moving persons. When using these low-resolution thermal images for facial emotion recognition this can negatively influence the emotion classification accuracy. To tackle the problem of low-resolution thermal facial expression images, super-resolution can be used. In this exploratory work, we propose the Thermal Face Super-Resolution Network (TFSRNet) and the Thermal Face Super-Resolution Generative Adversarial Network (TFSRGAN) to recover high-resolution thermal facial expression images from low-resolution thermal facial expression images, with the goal to use the super-resolved images for thermal facial emotion recognition. The architecture TFSRNet is optimized to minimize the mean squared error (MSE), which results in images with a high peak signal-to-noise ratio (PSNR). However, these images often contain an unsatisfying perceptual quality. To generate high-resolution images with a high perceptual quality we propose TFSRGAN. Both architectures use facial prior knowledge, such as facial landmark heatmaps and parsing maps, to enhance low-resolution thermal facial expression images. To emphasize the most important parts of each facial expression and to suppress irrelevant facial parts, we integrate the Convolutional Block Attention Module (CBAM) in both super-resolution architectures. The proposed super-resolution architectures are used to enhance low-resolution thermal facial expression images, which are obtained with three different degradation models, namely bi-cubic down-sampling (BI) on scale x2, x3 and x4, blurring followed by bi-cubic down-sampling (BD) on scale x3 and bi-cubic down-sampling followed by adding noise (DN) on scale x3. With an ablation study, the effectiveness of using facial prior knowledge and the attention mechanism CBAM for thermal super-resolution is shown. When using facial prior knowledge and the attention mechanism CBAM, the image quality of the super-resolved images improves. Furthermore, experiments show that images enhanced by TFSRNet outperform bi-cubic interpolated images, for degradation models BI x4, BD x3 and DN x3. Using these super-resolved images for thermal facial emotion recognition also leads to an increase of the emotion classification accuracy. In addition, images enhanced by TFSRGAN outperform bi-cubic interpolated images for degradation model DN x3. Although, this an exploratory work containing limitations, the experiments show the effectiveness of using facial prior knowledge and the attention mechanism CBAM for thermal facial expression super-resolution. In addition, thermal face super-resolution shows promising results for thermal facial emotion recognition where future work can build upon.

To create safe urban areas, it is important to gain insight into what influences the (perceived) safety of our cities and human settlements. One of the factors that can contribute to safety is the way urban spaces are designed. Previous work has highlighted the importance of natural surveillance: a type of surveillance that is a byproduct of how citizens normally and routinely use the environment. However, studying this concept is not a trivial task. Manual approaches such as observation studies are costly and time consuming and have therefore often limited themselves to smaller geographical areas.

In this work, we present a methodology that can automatically provide an estimate of natural surveillance by detecting building openings (i.e. windows and doors) in street level imagery and localizing them in 3 dimensions. The proposed method is able to estimate natural surveillance at the street segment level, while simultaneously being able to gather data on a whole city in a matter of hours. We then apply our method to the city of Amsterdam to analyze the relationship between natural surveillance and urban safety using the Amsterdam Safety Index.

We conclude that our chosen operationalization of natural surveillance (road surveillability and occupant surveillability) is correlated with decreases in high impact crime and nuisance as well as increases in perceived safety. Furthermore we provide evidence for the existence of a threshold after which extra natural surveillance is no longer associated with higher degrees of safety.

To study how to involve the end-users in the development of machine learning explainability, this project has chosen the context of bird species identification. It intends to develop a platform where the end-users can learn bird knowledge while contributing to building the explainability of machine learning models. Among all the methods that equips machine learning models with explainability, this project adopts a framework called SECA (Semantic Concept Extraction and Analysis). In this framework, we require human-made-annotations to be made to the saliency maps of training photos to provide semantically understandable explanations to the end-users. On the other hand, we hope that the process of making annotations will also benefit the human annotators’ skills in bird species identification, in order to motivate their participation. Two main goals of the user research were: to understand the users’ needs for learning and to know their capability in making the annotations needed by the project owners. The user research started with qualitative and quantitative research to understand the current practices of the bird hobbyists, to define the target user groups, which were the birders with zero or little expertise. Then, in order to link their learning needs to the capability of machine learning explanations, three prototypes were built to collect their feedback. It was found out that they didn’t care much about the justification or transparency of bird ID apps, compared to learning knowledge in distinguishing birds. Then came the annotation test when we found the participants were able to finish the annotation task with high correctness (>93% on average). And the most popular annotations of each task were 100% correct. Finally, we built a functional, high-fidelity prototype with experiential interfaces and interactions, and tested it among 3 of the target users. They had positive feedback on the prototype’s usability and the overall workflow, which proved the feasibility of our concepts. Recommendations on usability were drawn at the end of this test. Throughout the research and design phases in this project, we have developed an approach to involve end-users in the annotation process of an explainable bird species identification model for their own benefit of fun and learning, which could potentially be applied to broader deployments

Data-centric research is all about trying to obtain useful insights on products that its researchers trying to create/design by collecting usage data on their prototypes. This type of research, usually performed by data-scientist, make use of machine learning techniques in order to improve any system or product that makes use of automated decisions. Now that other researchers, in this case researchers from design studies, would like to perform data-centric research. Therefore, they must be made aware of the issues that could be introduced by using machine learning to automate decisions in electronic devices. To address the issues around bias, we thought that it would be good to make them aware through some type of system where they should be made aware of bias. So we proposed that there was a need for a system where the user could explore bias present in the dataset or introduced by machine learning practices. To be able to find biases we had to decide which bias toolkit which led to use of the AIF360 toolkit that we eventually used to create FairData. FairData lets the user explore bias in datasets that include machine learning predictions by going through and iterative process consisting of a exploration stage and a result stage. The exploration part contains the parts for analysing the datasets by going through the attributes and making small modifications to the dataset. At the results page of this system it will show a change in bias represented by multiple metrics to support the users in becoming aware of bias. The experimentation part of this thesis showed that our system is able to inform two different groups of participants (compute science & industrial design) about the bias present in the dataset. The experiment made it clear that even though participants have different behaviour when analysing/exploring the dataset, they all tend to become aware of the bias by looking at the different metrics provided by FairData. It also seems that it is possible to support users in trying to reduce this bias as they can see the effects of the modifications done to the dataset.

A significant portion of academic graduates have difficulty finding a first job after graduation. Research shows that the expectations of academic graduate job seekers and employers do not align and this graduation project confirms that job seekers and employers do not speak the same language. On the one hand, job seekers do not seem very able to communicate their skills and abilities in a convincing manner. On the other hand, employers do not seem very able to communicate the job requirements effectively. In this graduation project, Jeroen ter Haar Romenij validated and developed a vacancy platform in collaboration with the start-up HelloCareer. The platform allows academic graduate job seekers to explore job opportunities with the use of a job task language. With this task-language, HelloCareer aims to bridge the gap between educational study programs and actual job profiles on the labour market. The task language enables job seekers and employers to express their preferences, respectively for a future job and a future employee, in a uniform language, thereby reducing the asymmetry of information between the two parties, ultimately resulting in better matches. During this graduation project, a task-language for the three master programmes that are a part of the TU Delft’s faculty of Industrial Design Engineering was co-developed with academic graduate job seekers. Based on intense involvement of both the job seekers and the employers, crucial learnings were acquired on how to best define and apply the task-language and shape the design of the task-based vacancy platform. The way in which the preferences of the job seekers are represented by the platform has a direct effect on the job opportunities that are presented to them. Therefore, the value of autonomy over self-representation is highly at stake and has been put central in the development of the task-based vacancy platform. To engrain this way of thinking in the design process, a design for values approach has been chosen. Through empirical research that was conducted with academic job seekers, it has been explored what the value of autonomy over self-representation means for them in the context of the vacancy platform. As a result, these insights have shaped the design of the task-based vacancy platform which is described in this thesis. The final result is of this graduation project is a User Interface Design that demonstrates in a clear and practical manner how the task-based vacancy platform operates.

The Dutch version of the Public Records Request is named the ‘Wet Openbaarheid van Bestuur’ (Wob) , which provides the public with the right to request access to records from any governmental institution. The government has the obligation to provide information about policy and the execution of policy; however individuals who wish to obtain more detailed information, can request that a governmental body publicly disclose certain information. Certain steps that require large amounts of time continue to exists when processing such a request. One of these tasks is described in this thesis, namely the redaction of personal opinions within internal deliberations. The goal of this thesis was to investigate the possibility of automatically recognising personal opinions within internal deliberations in order to speed up the process of handling a Wob-request.

Daily stress is a problem that many people are suffering from. In previous research a prototype was developed to help users to identify their own stressful places and encourage them to conduct self-training exercises in those locations. This prototype is called ‘Grippy’. The goal of this project is to explore what qualities Grippy’s vibration signal should have, to appropriately warn the user of upcoming stress. Three knowledge gaps are identified. First, how the vibration strength of the signal influences the noticeability, audibility and disruption of the vibration signal. Second, what environmental factors influence the noticeability of wrist-mounted vibration signals. And third, how we could design respectful vibration signals that grab the users attention. Three experiments have been performed. Experiments one and two used 14 and 7 participants respectively to measure at which vibration strength Grippy’s vibration signals are noticeable to the user and bystanders respectively. With the insights from these two experiments, a new stress alert signal is proposed. This signal is tested in a third auto-introspective experiment. In this experiment we also explore ‘how Grippy fulfils the qualities of wearable partners in daily life’. The proposed stress alert signal is found to be discreet and respectful in most situations. These include social situations like presenting, listening and talking, but also shopping, walking, studying and playing piano. During cycling the proposed stress alert signal was not consistently noticeable. In addition, short disruptions in the environment such as arm movement and gusts of wind could temporarily distract from or mask the vibration signal. This report concludes that a continuous vibration signal at a fixed vibration strength is likely an appropriate way of alerting users in most situations, but that adaptation of the vibration strength to detect cycling like situations will be necessary.

People like to travel in groups to visit places. Group recommendation systems can be used to recommend an itinerary of "places of interests" (POIs) in an ordered sequence. The order of POIs in the sequence can be explained to group members to increase acceptance of the recommended items. There is a possibility that explanations which reveal names and rating preferences could create a threat to privacy. The main study in this work uses two group types - a primary group consisting of closely-related members, and a secondary group consisting of loosely-related members. Explanations with either complete information or privacy-preserving information are offered alternatively to these groups. The purpose of this study is to evaluate whether different group types need different types of explanations to improve their satisfaction. These explanations explain the entire recommended sequence of POIs with regard to possible conflicting situations that could occur due to disagreement with the order of the sequence. A total of 25 participants took part in the evaluation. There was no significant difference identified between the explanation types preferred by each group type. To understand the underlying reason for this result, a post-hoc analysis was done. We identified a participant's most frequently used conflict-handling modes using the Thomas-Kilmann personality assessment test. We then analyzed the user comments provided during the questionnaire. The analysis potentially suggests that different conflict-handling modes could be a factor affecting which explanation type was preferred by a person when they are in a particular group (e.g. primary vs secondary).

Log data, produced from every computer system and program, are widely used as source of valuable information to monitor and understand their behavior and their health. However, as large-scale systems generate a massive amount of log data every minute, it is impossible to detect the cause of system failure by examining manually this huge size of data. Thus, there is a need for an automated tool for finding system's failure with little or none human effort. Nowadays lots of methods exist that try to detect anomalies on system's logs by analyzing and applying various algorithms such as machine learning algorithms. However, experts argue that a system error can not be found by looking into a single event, but in multiple log event data are necessary to understand the root cause of a problem. In this thesis work, we aim to detect patterns in sequential distributed system's logs that can capture effectively the abnormal behavior. Specifically as a first step, we will apply rule mining techniques to extract rules that represent an anomalous behavior, which potentially in the future may lead to a failure of a system. Except for that step, we implemented a real-time anomaly detection framework to detect problems before they actually occur.
Processing log data as streams is the only way to achieve a real-time detection concept. In that direction we will process streaming log data using a complex event processing technique. Specifically, we would like to combine rule mining algorithms with complex event processing engine to raise alerts on abnormal log data based on automatically generated patterns. The evaluation of the work is conducted on Hadoop's logs, a widely used system in the industry. The outcome of this thesis project gives really promising results, reaching a Recall of 98\% in detecting anomalies. Finally, a scalable anomaly detection framework was build by integrating different systems into the cloud. The motivation behind this is the direct application of our framework to a real-life use case.

RFID technology is slowly replacing traditional bar codes as a way to identify and track objects and individuals. However, consumer-oriented market penetration has been limited as dedicated RFID readers carry a high start-up cost. Furthermore, interactions with individual tags require special-purpose RFID readers. We present a novel RFID tag interaction system based on commercially of the shelf hardware at vastly lower cost compared to conventional systems.

Web shops use recommender systems to help users find the products they find interesting in the large amount of available products online. An often used approach to do so is collaborative filtering. This method relies on historical user-item interactions and uses them to recommends products other users found interesting. Fashion is very reliant on quickly changing trends and personal preferences and requires a more personal and up-to-date approach. The focus of this research is to generate recommendations based on what products the user is currently searching for. It does this by detecting user behaviour based on the search scope of users and products user look at in the current session. Then new products are recommended by means of clustering new products to the most interesting products of the current session. This system was then compared with item-based collaborative filtering with an A/B test on the fashion platform Fashionchick.nl. It was found that traditional collaborative filtering was slightly more effective, but because of the small differences it is concluded that a behaviour driven recommender system are be promising and that more work is needed.

Locating people inside buildings is still an unsolved problem. There is a lot
of research going on in this field and many different solutions using different
techniques have been proposed. However, there is no widely accepted indoor
localization solution like how GPS is for outdoor localization due to less accuracy, higher hardware requirement, cost etc,. We introduce a system that locates
people indoors more accurately.

With the increasing use of big data in varied applications to improve decision making and provide new insights, the research explores the potential of the Uber Movement data set released by Uber comprising of travel times from one zone to the other. A better understanding of the potential of the dataset could lead to the addition of the existing tool kit of Transport planners and city officials at the municipality of Amsterdam. Moreover, it would be the first of a kind data set enabling an understanding of taxi movement in the city. The Uber Movement Travel Time comprises of the average travel time between two wijken, where the ‘sourceid’ and ‘dstid’ do not correspond to the origin and destination of a trip but simply represent the directionality of the travel time measured. The data is aggregated across different levels of temporal detail and the number of data points directly corresponds to the level of temporal aggregation. For instance, if the quarterly aggregated data for the different days of the week is downloaded, the number of data points between a ‘sourceid’ and ‘dstid’ cannot exceed seven.
Three aspects of the data set were explored: 1) ability to capture the demand for Ubers 2) ability to capture recurrent congestion and 3) ability to capture non-recurrent congestion. While the data according to the Uber Movement and previously used instances, the data is suited for performance (recurrent congestion and non-recurrent congestion) and impact-related studies of the network. The absence of route related information limits the applications of the data. The potential of the data is also limited by the data sparsity. The potential of the data was best revealed through demand studies which indicated a skewed user group of tourists, airport users (to and fro), work-related trips and users using Ubers late at night. In addition, with respect to the goals of the municipality in managing traffic activity across different zones and time periods, by implementing and extending an existing model in the form of adding ‘occupancy related measures’ and ‘shortest path’. Thus, based on the data penetration levels and travel time data, the model developed offers insights at a strategic level to the city in the form of Spatio-temporal concentration of Uber vehicles, occupancy levels through the day. The potential of the data lies in its ability to offer strategic insights to the city of Amsterdam and the greater Amsterdam region in the form of the unique Spatio-temporal spread of Uber vehicles across different hours of the day.

The accelerating growth of the Internet of Things (IoT) has led to the development of many different communication protocols to enable the most optimal environment for the nature of IoT. Requiring long range and low power communication abilities, has resulted in LPWANs (low-power WANs); one such LPWAN is LoRaWAN (Long Range WAN) which is studied in depth in this thesis.The thesis is done in collaboration with The Things Network (TTN), a crowd-sourced LoRaWAN expansing over 6 continents. The data retrieved from the TTN NOC (Network Operations Centre) has played a crucial role in this thesis, as it provides the basis for studying the working and performance of a LoRaWAN.The aim of this study is to use performance evaluation of the network to develop an Adaptive Data Rate (ADR) model which will modify the transmission parameters to improve the performance of the network, in terms of the ratio of received and sent packets (called Data Extraction Rate - DER), while attempting to maintain minimal cost of transmission, in terms of transmission power and usage of bandwidth by measuring the airtime used in transmission.Extensive evaluation and analysis of the NOC data is performed and detailed in this report, followed by a modified ADR model which theoretically will improve the DER of the network. This ADR model is verified by modelling the performance of devices on the network with the usage of the proposed ADR model and without. This theoretical verification proves that the DER of the network improves when the transmission parameters are varied in accordance with the proposed ADR model.

In the last years, the popularity of video-on-demand services has been constantly increasing, especially for the young audiences who are more adept at using new technologies. Through those platforms, the viewers have access to a huge volume of movies at any moment that makes the viewing decision for most of them a very challenging task. Recommender systems are employed by video-on-demand providers to address the former challenge. We propose a novel movie recommender system that filters movies based on the genre-related visual elements of their trailers. The proposed system utilizes a 3D pre-trained deep ConvNet to extract spatio-temporal deep features from the trailers which then are combined, through a Deep Bag of Segments (DBoS) pooling network, with the genre information of the movie to provide a single movie representation. The 3D deep visual genre-aware representation is exploited by a pure content-based filtering system to provide personalized recommendations to users. We conduct offline experiments with two datasets to evaluate the performance of our approach with respect to accuracy and beyond accuracy metrics. We also conduct an online experiment in a real-world streaming platform to evaluate the user perceived utility of the recommendations produced by a pure content-based recommender system using our proposed genre-aware movie descriptor against the same system using genre and visual 3D deep features. We conclude that a continuous genre representation, which reflects genre specific visual elements of the movie, provides interesting results in the content-based movie recommendation task. Exploring further its potential could bring important benefits to various tasks in the movie domain.

Progress has been made recently in the research of equal hierarchy backscatter communications. There are, however, challenges to overcome before this technology can become mainstream. For example, state-of-the-art solutions to the phase cancellation problem that occurs in such networks are still trivial and sub-optimal, especially in terms of resource utilization. This thesis aims to address the issue by designing a novel networking algorithm. This is done by (i) characterizing and analyzing such networks to assess the potential of the proposal, and (ii) experimentally demonstrating the concept in the largest implemented network so far. Findings show that the solution presented in this work not only matches current phase cancellation solutions in terms of robustness, but also extends communication range and enables connecting networks served by separate exciters.

Being capable to foresee the future of a given financial asset as an investor, may lead to significant economic profits. Therefore, stock market prediction is a field that has been extensively developed by numerous researchers and companies. Recently, however, a new branch of financial assets has emerged, namely cryptocurrencies. As a representative of these tokens, we chose the largest and most popular cryptocurrency, called Bitcoin. Its value is characterized by with non-stationary behaviour and occurrence of speculative bubbles, which cause a rapid explosion of the price, followed by a major crisis and market panic.

Currently, most of the research community does not take these issues into account, while predicting its price, which may lead to wrong conclusions or unstable results. Therefore, in this thesis, we take a step back and reconsider how does the environment influence model's performance and how to use this knowledge to implement more accurate forecast in the future. Moreover, by designing an appropriate methodology and employing semantic features from online text sources, such as Twitter, Reddit and online news portals, we attempt to build a robust prediction system that offers stable performance regardless of the market fluctuations.

Executed experiments prove that non-stationarity negatively influences the results, causing the deterioration of model's performance over time. Furthermore, it appears that there may be certain properties of economic bubbles that facilitate more efficient prediction, as well as some predictors have an ability to successfully forecast the beginning of a market crisis. However, these findings are based on individual observations, which need to be confirmed by further research. In addition, by designing an appropriate methodology, we prevented performance deterioration, caused by price signal non-stationarity. Although, the semantic features based on online sources did not boost the robustness of the system significantly, combined with the suitable system's design, they lead to improvement in the overall performance of the predictor.