Pub Date : 2019-04-01DOI: 10.1109/SIEDS.2019.8735617
Stephen Alexander, Juan S. Rozo, Bianca Donadio, N. Tenhundfeld, E. D. de Visser, Chad C. Tossell
The U.S. Air Force (USAF) mission planning process, briefing, and debriefing are critical for tactical planning based on operational objectives and team performance. Our research focuses on modernizing these processes through the use of new technology, which is tailored to the current and developing operational capabilities of the Department of Defense. The current state of mission planning in the USAF revolves around traditional methods, such as briefing missions with white boards, topographic maps, and toy planes. At the same time, missions and technologies are becoming increasingly complex. Although the methods in which operations are being conducted are advancing, methods to prepare for these operations are not seeing the same progress and it is vital to leverage new techniques to facilitate the efficient planning of complex missions. Our goal is to revolutionize the mission planning process by integrating modern technology into the process. To accomplish this, we looked to enhance mission planning, briefing, and debriefing with a more visual and immersive experience. We created a system in which “mission critical” intelligence data can be placed into an Excel spreadsheet and transformed into a virtual representation of the mission that can be viewed through virtual reality and augmented reality platforms. This implementation of present-day technology into mission planning operations has the potential to enhance mission performance and therefore reduce potential equipment damage and loss, as well as casualties.
{"title":"Transforming the Air Force Mission Planning Process with Virtual and Augmented Reality","authors":"Stephen Alexander, Juan S. Rozo, Bianca Donadio, N. Tenhundfeld, E. D. de Visser, Chad C. Tossell","doi":"10.1109/SIEDS.2019.8735617","DOIUrl":"https://doi.org/10.1109/SIEDS.2019.8735617","url":null,"abstract":"The U.S. Air Force (USAF) mission planning process, briefing, and debriefing are critical for tactical planning based on operational objectives and team performance. Our research focuses on modernizing these processes through the use of new technology, which is tailored to the current and developing operational capabilities of the Department of Defense. The current state of mission planning in the USAF revolves around traditional methods, such as briefing missions with white boards, topographic maps, and toy planes. At the same time, missions and technologies are becoming increasingly complex. Although the methods in which operations are being conducted are advancing, methods to prepare for these operations are not seeing the same progress and it is vital to leverage new techniques to facilitate the efficient planning of complex missions. Our goal is to revolutionize the mission planning process by integrating modern technology into the process. To accomplish this, we looked to enhance mission planning, briefing, and debriefing with a more visual and immersive experience. We created a system in which “mission critical” intelligence data can be placed into an Excel spreadsheet and transformed into a virtual representation of the mission that can be viewed through virtual reality and augmented reality platforms. This implementation of present-day technology into mission planning operations has the potential to enhance mission performance and therefore reduce potential equipment damage and loss, as well as casualties.","PeriodicalId":265421,"journal":{"name":"2019 Systems and Information Engineering Design Symposium (SIEDS)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2019-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123348474","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2019-04-01DOI: 10.1109/SIEDS.2019.8735628
Elizabeth Pennington, R. Hafer, Erin Nistler, Todd Seech, Chad C. Tossell
As virtual reality and artificial intelligence technologies continue to advance, the United States Military is quickly integrating these capabilities into initial flight training through efforts like the Air Force's Pilot Training Next (PTN) program. A persistent issue, however, has been a lack of data guiding (1) the ideal degree of integration into traditional pilot training and (2) the optimal amount of structure for student pilots' training experience. The goal of this study was to evaluate the aforementioned PTN model when applied to the U.S. Air Force Academy's flight training program with special emphasis on the ideal degree of structure for airmanship success. To this end, a quasi-experimental approach was utilized, which included 60 USAFA cadets enrolled in the Powered Flight Program who were pseudo-randomly assigned to three independent groups with varying degrees of structure. The groups (i.e., High Structured, Scaffolded, and Low Structured Groups) represented a spectrum of VR-training curriculum structure ranging from a rigid, lineal objective-completion model (akin to traditional flight training) to an unguided, Montessori-like model. With group assignment as the independent variable, live-flight performance was used as the dependent variable, which was quantified using flight grade cards, number of “landing tabs” (i.e., modified solos) awarded, and a subjective Instructor Pilot rating. Subjective feedback was also obtained from students in each condition. Initial effectiveness data indicated an increased level of perceived self-efficacy in coordination with increased virtual reality simulator time as well as an accelerated rate of positive transfer to real aircraft from the strictly structured and scaffolded groups. The results of this study allow for initial recommendations for forthcoming airmanship training and undergraduate pilot training augmentation efforts across the Department of Defense.
{"title":"Integration of Advanced Technology in Initial Flight Training","authors":"Elizabeth Pennington, R. Hafer, Erin Nistler, Todd Seech, Chad C. Tossell","doi":"10.1109/SIEDS.2019.8735628","DOIUrl":"https://doi.org/10.1109/SIEDS.2019.8735628","url":null,"abstract":"As virtual reality and artificial intelligence technologies continue to advance, the United States Military is quickly integrating these capabilities into initial flight training through efforts like the Air Force's Pilot Training Next (PTN) program. A persistent issue, however, has been a lack of data guiding (1) the ideal degree of integration into traditional pilot training and (2) the optimal amount of structure for student pilots' training experience. The goal of this study was to evaluate the aforementioned PTN model when applied to the U.S. Air Force Academy's flight training program with special emphasis on the ideal degree of structure for airmanship success. To this end, a quasi-experimental approach was utilized, which included 60 USAFA cadets enrolled in the Powered Flight Program who were pseudo-randomly assigned to three independent groups with varying degrees of structure. The groups (i.e., High Structured, Scaffolded, and Low Structured Groups) represented a spectrum of VR-training curriculum structure ranging from a rigid, lineal objective-completion model (akin to traditional flight training) to an unguided, Montessori-like model. With group assignment as the independent variable, live-flight performance was used as the dependent variable, which was quantified using flight grade cards, number of “landing tabs” (i.e., modified solos) awarded, and a subjective Instructor Pilot rating. Subjective feedback was also obtained from students in each condition. Initial effectiveness data indicated an increased level of perceived self-efficacy in coordination with increased virtual reality simulator time as well as an accelerated rate of positive transfer to real aircraft from the strictly structured and scaffolded groups. The results of this study allow for initial recommendations for forthcoming airmanship training and undergraduate pilot training augmentation efforts across the Department of Defense.","PeriodicalId":265421,"journal":{"name":"2019 Systems and Information Engineering Design Symposium (SIEDS)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2019-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116303319","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2019-04-01DOI: 10.1109/SIEDS.2019.8735634
E. Drummond, P. Condro, Ben Cotton, C. Cox, A. Pinegar, Kyle Vickery, R. Prins
Engineering students at James Madison University are creating an all-electric motorcycle as part of a two-year capstone design project. The final product will be an educational system that promotes access to the electric powertrain (consisting of tractive battery pack, battery management system (BMS), motor controller, and motor). This paper focuses on development of system performance parameters, design of major components including chassis and battery pack/BMS enclosure, and signal interactions between powertrain components. Previous iterations of electric motorcycle conversions developed at JMU were constrained by the donor chassis which were designed for support of internal combustion engines. Although teams worked to optimize the fitment of powertrain components within existing frame members, compromises were necessary. Other limitations of previous iterations include battery pack discharge rates and delicate battery management systems. Although electric motorcycles are commercially available, their powertrain components are generally proprietary and inaccessible (not available for hacking or other educationally appropriate activities). The current iteration was developed to address these limitations. Results include benchmarking results, estimation of performance, and physical iterations of design choices. The final iteration of the modular battery pack, designed for student interaction, consists of seven sub-pack modules with visible and intuitive wire routing. The completed powertrain is designed to favor accessibility of components as well as optimize available space within the frame while closely matching the center of gravity and suspension as well as steering capabilities of the donor motorcycle.
{"title":"Design and Construction of an Electric Motorcycle","authors":"E. Drummond, P. Condro, Ben Cotton, C. Cox, A. Pinegar, Kyle Vickery, R. Prins","doi":"10.1109/SIEDS.2019.8735634","DOIUrl":"https://doi.org/10.1109/SIEDS.2019.8735634","url":null,"abstract":"Engineering students at James Madison University are creating an all-electric motorcycle as part of a two-year capstone design project. The final product will be an educational system that promotes access to the electric powertrain (consisting of tractive battery pack, battery management system (BMS), motor controller, and motor). This paper focuses on development of system performance parameters, design of major components including chassis and battery pack/BMS enclosure, and signal interactions between powertrain components. Previous iterations of electric motorcycle conversions developed at JMU were constrained by the donor chassis which were designed for support of internal combustion engines. Although teams worked to optimize the fitment of powertrain components within existing frame members, compromises were necessary. Other limitations of previous iterations include battery pack discharge rates and delicate battery management systems. Although electric motorcycles are commercially available, their powertrain components are generally proprietary and inaccessible (not available for hacking or other educationally appropriate activities). The current iteration was developed to address these limitations. Results include benchmarking results, estimation of performance, and physical iterations of design choices. The final iteration of the modular battery pack, designed for student interaction, consists of seven sub-pack modules with visible and intuitive wire routing. The completed powertrain is designed to favor accessibility of components as well as optimize available space within the frame while closely matching the center of gravity and suspension as well as steering capabilities of the donor motorcycle.","PeriodicalId":265421,"journal":{"name":"2019 Systems and Information Engineering Design Symposium (SIEDS)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2019-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127567666","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2019-04-01DOI: 10.1109/SIEDS.2019.8735611
Gage Beckwith, Tim Callahan, Bear Carlson, Tyler Fondren, R. Harris, Jacqueline Hoege, Tykai Martin, Collin Menna, Ella Summer, W. Scherer, Chris Tuttle, Stephen Adams
The role that data analytics plays on sports teams has increased dramatically since Michael Lewis wrote Moneyball and shed some light on Billy Beane's use of analytics with the Oakland Athletics. Today, every major professional sports team has at least an analytics expert on staff, if not a whole department [1]. College teams are increasing their use of analytics as well. Our research goals were to improve the University of Virginia (U. Va.) football team in two ways: recruiting and on-field performance. Our goal of improving the recruiting process led to the development of two tools. First, we created a model that predicts how well an athlete will perform in college based on their high school statistics and demographics. This tool allows coaches to discover lesser ranked athletes who are likely to outperform their rankings. We also further developed an existing model that predicts how likely players are to commit to U. Va. This tool prevents coaches from potentially wasting valuable time and resources on players who are unlikely to commit to U. Va. In order to improve U. Va.'s on-field performance, we created two additional tools. We developed an expected points model based on existing NFL models in an attempt to evaluate the team's performance and identify areas where our play calling was consistently sub-optimal. Finally, we created matchup reports that the coaches can use to scout opposing teams. The expected points model is integrated into these reports to provide a more accurate assessment of the opponent's performance. With this tool, the coaches will be able to spend less time identifying opponents' strengths and weaknesses and more time preparing to exploit them.
{"title":"Systems Analysis for University of Virginia Football Recruiting and Performance","authors":"Gage Beckwith, Tim Callahan, Bear Carlson, Tyler Fondren, R. Harris, Jacqueline Hoege, Tykai Martin, Collin Menna, Ella Summer, W. Scherer, Chris Tuttle, Stephen Adams","doi":"10.1109/SIEDS.2019.8735611","DOIUrl":"https://doi.org/10.1109/SIEDS.2019.8735611","url":null,"abstract":"The role that data analytics plays on sports teams has increased dramatically since Michael Lewis wrote Moneyball and shed some light on Billy Beane's use of analytics with the Oakland Athletics. Today, every major professional sports team has at least an analytics expert on staff, if not a whole department [1]. College teams are increasing their use of analytics as well. Our research goals were to improve the University of Virginia (U. Va.) football team in two ways: recruiting and on-field performance. Our goal of improving the recruiting process led to the development of two tools. First, we created a model that predicts how well an athlete will perform in college based on their high school statistics and demographics. This tool allows coaches to discover lesser ranked athletes who are likely to outperform their rankings. We also further developed an existing model that predicts how likely players are to commit to U. Va. This tool prevents coaches from potentially wasting valuable time and resources on players who are unlikely to commit to U. Va. In order to improve U. Va.'s on-field performance, we created two additional tools. We developed an expected points model based on existing NFL models in an attempt to evaluate the team's performance and identify areas where our play calling was consistently sub-optimal. Finally, we created matchup reports that the coaches can use to scout opposing teams. The expected points model is integrated into these reports to provide a more accurate assessment of the opponent's performance. With this tool, the coaches will be able to spend less time identifying opponents' strengths and weaknesses and more time preparing to exploit them.","PeriodicalId":265421,"journal":{"name":"2019 Systems and Information Engineering Design Symposium (SIEDS)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2019-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124967418","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2019-04-01DOI: 10.1109/SIEDS.2019.8735597
Utkarsha Bhave, Grant D Showalter, Dalton J Anderson, Cesar Roucco, Andrew C Hensley, G. Lewin
The United States Department of Defense anticipates unmanned systems will be integrated into most defense operations by 2030 to reduce risk to human life, enhance reliability, and ensure operation consistency and efficiency. However, current technology requires human operation for ethical decision-making, leaving an opportunity to automate some tasks to assist operators. Previously, a University of Virginia capstone team designed an unmanned ground vehicle (“the rover”) to aid intelligence, surveillance, and reconnaissance missions in adversarial environments. However, a lack of GPS connectivity indoors and system latency limited the rover's performance and created a lag in the operator's view compared to the rover's true position, occasionally causing the operator to inadvertently crash the rover into obstacles. The objectives of this project are to mitigate operational risks by equipping the rover with functionalities to autonomously avoid obstacles, map an unknown indoor space, and navigate itself back to a predetermined location (“base”). Obstacle avoidance is accomplished through an algorithm that stops the rover a safe distance away from a detected obstacle, but still allows the human operator to navigate the rover away from the obstacle prior to continuing the mission. Algorithms are implemented to perform Simultaneous Localization And Mapping and to determine best-route navigation to the base. Laser rangefinder data, an improved processor, and, potentially, visual odometry sensors are used to aid in the navigation algorithms. Testing has confirmed that the rover successfully stops in front of laser-detected obstacles, builds digital maps of an unknown indoor space, and can navigate back to a base, though the performance has room for improvement. It is anticipated that incorporating visual odometry can enhance the rover's mapping implementation and obstacle avoidance performance.
{"title":"Automating the Operation of a 3D-Printed Unmanned Ground Vehicle in Indoor Environments","authors":"Utkarsha Bhave, Grant D Showalter, Dalton J Anderson, Cesar Roucco, Andrew C Hensley, G. Lewin","doi":"10.1109/SIEDS.2019.8735597","DOIUrl":"https://doi.org/10.1109/SIEDS.2019.8735597","url":null,"abstract":"The United States Department of Defense anticipates unmanned systems will be integrated into most defense operations by 2030 to reduce risk to human life, enhance reliability, and ensure operation consistency and efficiency. However, current technology requires human operation for ethical decision-making, leaving an opportunity to automate some tasks to assist operators. Previously, a University of Virginia capstone team designed an unmanned ground vehicle (“the rover”) to aid intelligence, surveillance, and reconnaissance missions in adversarial environments. However, a lack of GPS connectivity indoors and system latency limited the rover's performance and created a lag in the operator's view compared to the rover's true position, occasionally causing the operator to inadvertently crash the rover into obstacles. The objectives of this project are to mitigate operational risks by equipping the rover with functionalities to autonomously avoid obstacles, map an unknown indoor space, and navigate itself back to a predetermined location (“base”). Obstacle avoidance is accomplished through an algorithm that stops the rover a safe distance away from a detected obstacle, but still allows the human operator to navigate the rover away from the obstacle prior to continuing the mission. Algorithms are implemented to perform Simultaneous Localization And Mapping and to determine best-route navigation to the base. Laser rangefinder data, an improved processor, and, potentially, visual odometry sensors are used to aid in the navigation algorithms. Testing has confirmed that the rover successfully stops in front of laser-detected obstacles, builds digital maps of an unknown indoor space, and can navigate back to a base, though the performance has room for improvement. It is anticipated that incorporating visual odometry can enhance the rover's mapping implementation and obstacle avoidance performance.","PeriodicalId":265421,"journal":{"name":"2019 Systems and Information Engineering Design Symposium (SIEDS)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2019-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129840740","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2019-04-01DOI: 10.1109/SIEDS.2019.8735612
C. Morton, M. Mumford, N. Peterson, Ashlie Veronie, Heather Kirkvold
Patients of Van Nes Rotationplasty often experience pain in their residual limb, due to the novel nature of the surgery. Literature review reveals that this pain occurs in a location on the limb that coincides with an abnormal concentration of force with respect to normal loading conditions. This paper discusses the results of a project where common engineering techniques were used to redesign a prosthetic leg and alleviate this pain for a client. The scope, for the purpose of this process, has been limited to the prosthetic “socket” where the residual limb sits. Use of 3D modeling and printing allows for quick, low cost iteration of the socket for testing, and is thus critical to the process. In full, the paper provides a prescriptive method to redesign a rotationplasty prosthesis towards the same result, for any client. The developed methodology for testing the device utilizes force sensors placed inside the socket, comparing the internal forces between a new model and an original, problematic one. In addition to this force measurement method, the process implements a prosthetic comfort evaluation form, allowing a client to qualitatively provide feedback on a prosthetic. Existing literature implied that reduction of force at the location of observed pain will reduce that pain, and client testing confirmed that notion, confirming the viability of the force sensor testing method.
{"title":"Redesigning a Rotationplasty Prosthetic","authors":"C. Morton, M. Mumford, N. Peterson, Ashlie Veronie, Heather Kirkvold","doi":"10.1109/SIEDS.2019.8735612","DOIUrl":"https://doi.org/10.1109/SIEDS.2019.8735612","url":null,"abstract":"Patients of Van Nes Rotationplasty often experience pain in their residual limb, due to the novel nature of the surgery. Literature review reveals that this pain occurs in a location on the limb that coincides with an abnormal concentration of force with respect to normal loading conditions. This paper discusses the results of a project where common engineering techniques were used to redesign a prosthetic leg and alleviate this pain for a client. The scope, for the purpose of this process, has been limited to the prosthetic “socket” where the residual limb sits. Use of 3D modeling and printing allows for quick, low cost iteration of the socket for testing, and is thus critical to the process. In full, the paper provides a prescriptive method to redesign a rotationplasty prosthesis towards the same result, for any client. The developed methodology for testing the device utilizes force sensors placed inside the socket, comparing the internal forces between a new model and an original, problematic one. In addition to this force measurement method, the process implements a prosthetic comfort evaluation form, allowing a client to qualitatively provide feedback on a prosthetic. Existing literature implied that reduction of force at the location of observed pain will reduce that pain, and client testing confirmed that notion, confirming the viability of the force sensor testing method.","PeriodicalId":265421,"journal":{"name":"2019 Systems and Information Engineering Design Symposium (SIEDS)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2019-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130569571","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2019-04-01DOI: 10.1109/SIEDS.2019.8735602
Justin Niestroy, Jiangxue Han, Jingyi Luo, Runhao Zhao, D. Lake, A. Flower
This study focuses on detecting deterioration of acutely ill patients in the cardiac ward at the University of Virginia Health System. Patients in the cardiac ward are expected to recover from a variety of cardiovascular procedures, but roughly 5% of patients deteriorate and have to be transferred to the Intensive Care Unit (ICU). Previous work has shown that early warning scores utilizing vitals signs and common lab results greatly lower morality for high risk patients. To build upon these results, data were collected over the course of two years from 71 beds in three cardiac-related wards at the University of Virginia Health System. In addition to information commonly collected for early warning scores, these data also contained continuous electrocardiography (ECG) telemetry data for all patients. Given that only one percent of observations are labeled as events, the F1 score was used as the primary metric to assess the performance of each model; area under the curve (AUC) was also considered. Previous work includes the development of logistic regression models with these data resulting in an AUC of 0.73. In this work, a super learner was built to further the study by stacking logistic regression, random forest, and gradient boosting models. Furthermore, a denoising auto-encoder was created to generate computer-derived features, the results of which were fed to machine learning models mentioned previously to predict patient deterioration. The logistic regression model built on existing and computer-generated features resulted in an F1 score of 0.1 and AUC of 0.7, which is comparable to previous models built on the same patient data set. The super learner had an improvement over existing logistic regression models, with an F1 score of 0.24 and AUC of 0.79.
{"title":"Prediction of Decompensation in Patients in the Cardiac Ward","authors":"Justin Niestroy, Jiangxue Han, Jingyi Luo, Runhao Zhao, D. Lake, A. Flower","doi":"10.1109/SIEDS.2019.8735602","DOIUrl":"https://doi.org/10.1109/SIEDS.2019.8735602","url":null,"abstract":"This study focuses on detecting deterioration of acutely ill patients in the cardiac ward at the University of Virginia Health System. Patients in the cardiac ward are expected to recover from a variety of cardiovascular procedures, but roughly 5% of patients deteriorate and have to be transferred to the Intensive Care Unit (ICU). Previous work has shown that early warning scores utilizing vitals signs and common lab results greatly lower morality for high risk patients. To build upon these results, data were collected over the course of two years from 71 beds in three cardiac-related wards at the University of Virginia Health System. In addition to information commonly collected for early warning scores, these data also contained continuous electrocardiography (ECG) telemetry data for all patients. Given that only one percent of observations are labeled as events, the F1 score was used as the primary metric to assess the performance of each model; area under the curve (AUC) was also considered. Previous work includes the development of logistic regression models with these data resulting in an AUC of 0.73. In this work, a super learner was built to further the study by stacking logistic regression, random forest, and gradient boosting models. Furthermore, a denoising auto-encoder was created to generate computer-derived features, the results of which were fed to machine learning models mentioned previously to predict patient deterioration. The logistic regression model built on existing and computer-generated features resulted in an F1 score of 0.1 and AUC of 0.7, which is comparable to previous models built on the same patient data set. The super learner had an improvement over existing logistic regression models, with an F1 score of 0.24 and AUC of 0.79.","PeriodicalId":265421,"journal":{"name":"2019 Systems and Information Engineering Design Symposium (SIEDS)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2019-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130250470","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2019-04-01DOI: 10.1109/SIEDS.2019.8735587
Nicola Bruno, Tommy Jun, Henry Tessier
The Logistics Management Institute (LMI) works with the US Department of Defense (DoD) in analyzing maintenance logs on US weapons systems. A major issue in processing this data is determining how to extract useful information from disorganized short-form texts in order to optimize the maintenance of these systems. Unlike text from other corpora, these text entries are only a few words in length and do not conform to lexical convention. LMI has provided a subset of about 10 million of these maintenance logs, each labeled with action-object pairs. The goals of this research are to construct a model that predicts action-object pairs and provide a metric to assess its validity. Prior to analysis, the entries are vectorized by either TFIDF and TSVD, or Word2vec. Several models are applied, including logistic regression, k-NN, SVM, decision trees, LSA, and DBSCAN clustering. Unsupervised models are tested in addition to supervised models due to the ambiguity regarding the validity of the provided ground truth values. The results of these tests yield accuracy scores of about 0.53 for action words and 0.73 for object words. Furthermore, the results from clustering provides evidence for discrepancies in the ground truth values. Taking this into consideration, prior models are adjusted and accuracy scores increased to 0.78 for action words.
{"title":"Natural Language Processing and Classification Methods for the Maintenance and Optimization of US Weapon Systems","authors":"Nicola Bruno, Tommy Jun, Henry Tessier","doi":"10.1109/SIEDS.2019.8735587","DOIUrl":"https://doi.org/10.1109/SIEDS.2019.8735587","url":null,"abstract":"The Logistics Management Institute (LMI) works with the US Department of Defense (DoD) in analyzing maintenance logs on US weapons systems. A major issue in processing this data is determining how to extract useful information from disorganized short-form texts in order to optimize the maintenance of these systems. Unlike text from other corpora, these text entries are only a few words in length and do not conform to lexical convention. LMI has provided a subset of about 10 million of these maintenance logs, each labeled with action-object pairs. The goals of this research are to construct a model that predicts action-object pairs and provide a metric to assess its validity. Prior to analysis, the entries are vectorized by either TFIDF and TSVD, or Word2vec. Several models are applied, including logistic regression, k-NN, SVM, decision trees, LSA, and DBSCAN clustering. Unsupervised models are tested in addition to supervised models due to the ambiguity regarding the validity of the provided ground truth values. The results of these tests yield accuracy scores of about 0.53 for action words and 0.73 for object words. Furthermore, the results from clustering provides evidence for discrepancies in the ground truth values. Taking this into consideration, prior models are adjusted and accuracy scores increased to 0.78 for action words.","PeriodicalId":265421,"journal":{"name":"2019 Systems and Information Engineering Design Symposium (SIEDS)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2019-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114213893","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2019-04-01DOI: 10.1109/SIEDS.2019.8735629
Ravi K U Rakesh, Boda Ye, D. Roden, Catherine Beazley, Karan Gadiya, Brendan Abraham, Donald E. Brown, M. Veeraraghavan
The volume of cyberattacks has grown exponentially over the last half-decade and shows no signs of slowing down. Additionally, attacks are rapidly evolving and are becoming increasingly more sophisticated. Cyber companies and academics alike have turned to machine learning to build models that learn data-driven rules for threat detection. However, these methods require a substantial amount of training data, and many enterprises lack the infrastructure to label their own network traffic for supervised learning. An added complexity to the labeling problem is that IP addresses are frequently reassigned to new hosts. In this paper, we lay a foundation for an autonomous traffic labeling pipeline that incorporates three different sources of ground truth and requires minimal manual intervention. We apply the labeling pipeline to network traffic data acquired from the University of Virginia. We process the network traffic with a popular network monitoring framework called Zeek, which provides aggregated statistics about the packets exchanged between a source and destination over a certain time interval. Additionally, the labeling pipeline synthesizes data from a network of honeypots compiled by the Duke STINGAR project, a series of nine blacklists, and a whitelist called Cisco Umbrella. We show, using cluster, port, and IP-location analyses, that a labeling methodology that ensembles the different data sources is better than one using only the individual sources. The labeling methodology proposed in the paper will aid enterprise network administrators in building robust intrusion detection systems.
{"title":"An Autonomous Labeling Pipeline for Intrusion Detection on Enterprise Networks","authors":"Ravi K U Rakesh, Boda Ye, D. Roden, Catherine Beazley, Karan Gadiya, Brendan Abraham, Donald E. Brown, M. Veeraraghavan","doi":"10.1109/SIEDS.2019.8735629","DOIUrl":"https://doi.org/10.1109/SIEDS.2019.8735629","url":null,"abstract":"The volume of cyberattacks has grown exponentially over the last half-decade and shows no signs of slowing down. Additionally, attacks are rapidly evolving and are becoming increasingly more sophisticated. Cyber companies and academics alike have turned to machine learning to build models that learn data-driven rules for threat detection. However, these methods require a substantial amount of training data, and many enterprises lack the infrastructure to label their own network traffic for supervised learning. An added complexity to the labeling problem is that IP addresses are frequently reassigned to new hosts. In this paper, we lay a foundation for an autonomous traffic labeling pipeline that incorporates three different sources of ground truth and requires minimal manual intervention. We apply the labeling pipeline to network traffic data acquired from the University of Virginia. We process the network traffic with a popular network monitoring framework called Zeek, which provides aggregated statistics about the packets exchanged between a source and destination over a certain time interval. Additionally, the labeling pipeline synthesizes data from a network of honeypots compiled by the Duke STINGAR project, a series of nine blacklists, and a whitelist called Cisco Umbrella. We show, using cluster, port, and IP-location analyses, that a labeling methodology that ensembles the different data sources is better than one using only the individual sources. The labeling methodology proposed in the paper will aid enterprise network administrators in building robust intrusion detection systems.","PeriodicalId":265421,"journal":{"name":"2019 Systems and Information Engineering Design Symposium (SIEDS)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2019-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124812594","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2019-04-01DOI: 10.1109/SIEDS.2019.8735638
Stephen V. Mortensen, Michael Christison, Bochao Li, AiLun Zhu, R. Venkatesan
The objectives of this project are two-fold: 1) to use statistical modeling techniques to help a Fortune 500 paper and packaging company codify what drives sales success and 2) to develop a model that can predict sales success with a reasonable degree of accuracy. The desired long-run result is to enable the company to improve both top-line revenue and bottom-line profits by increasing sales close rates, shortening sales cycles, and decreasing the cost of sales. The research team generated several models to predict win propensities for individual sales opportunities, choosing the model with the greatest predictive power and ability to generate insights to use as the backbone for a client tool. To accomplish this, the team leveraged structured and unstructured data from the company's Salesforce.com customer relationship management system. The team experimented with several techniques including binomial logit and various decision tree methods, including boosting with gradient boost and random forest. Individual attributes of customers, opportunities, and internal documentation methods that have the greatest influence on sales success were identified. The best model predicted win propensity with an accuracy of 80%, with precision and recall of 86% and 77%, respectively, which proved to be an improvement over current sales forecast accuracy.
{"title":"Predicting and Defining B2B Sales Success with Machine Learning","authors":"Stephen V. Mortensen, Michael Christison, Bochao Li, AiLun Zhu, R. Venkatesan","doi":"10.1109/SIEDS.2019.8735638","DOIUrl":"https://doi.org/10.1109/SIEDS.2019.8735638","url":null,"abstract":"The objectives of this project are two-fold: 1) to use statistical modeling techniques to help a Fortune 500 paper and packaging company codify what drives sales success and 2) to develop a model that can predict sales success with a reasonable degree of accuracy. The desired long-run result is to enable the company to improve both top-line revenue and bottom-line profits by increasing sales close rates, shortening sales cycles, and decreasing the cost of sales. The research team generated several models to predict win propensities for individual sales opportunities, choosing the model with the greatest predictive power and ability to generate insights to use as the backbone for a client tool. To accomplish this, the team leveraged structured and unstructured data from the company's Salesforce.com customer relationship management system. The team experimented with several techniques including binomial logit and various decision tree methods, including boosting with gradient boost and random forest. Individual attributes of customers, opportunities, and internal documentation methods that have the greatest influence on sales success were identified. The best model predicted win propensity with an accuracy of 80%, with precision and recall of 86% and 77%, respectively, which proved to be an improvement over current sales forecast accuracy.","PeriodicalId":265421,"journal":{"name":"2019 Systems and Information Engineering Design Symposium (SIEDS)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2019-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122889609","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：