Application-level Data Research Articles

Association Between a Medical School Applicant's Community College Attendance and the Likelihood of Application Acceptance: An Investigation of Select Medical School Characteristics.

One-third of medical school applicants attend a community college (CC), and they represent a diverse group of applicants. However, they have a lower likelihood of being accepted to medical school. Using application-level data, this study examines how an applicant's CC attendance impacts the likelihood of application acceptance and how 3 medical school characteristics moderate this association. Data examined were from 2,179,483 applications submitted to at least one of 146 U.S. Liaison Committee on Medical Education-accredited medical schools by 124,862 applicants between 2018 and 2020. The outcome was application acceptance. The main measures were applicants' CC attendance (no, lower [> 0%-19% of college course hours], or higher [≥ 20%]) and 3 medical school characteristics: geographic region, private versus public control, and admissions policy regarding CC coursework. Multilevel logistic regression models estimated the association between CC attendance and the likelihood of application acceptance and how this association was moderated by school characteristics. Among applicants, 23.8% (29,704/124,862) had lower CC attendance and 10.3% (12,819/124,862) had higher CC attendance. Regression results showed that, relative to no CC attendance, applications with lower (adjusted odds ratio [AOR] = .96; 95% confidence internal [CI], .94-.97) and higher (AOR = .78; 95% CI, .76-.81) CC attendance had significantly decreased odds of acceptance when the 3 school characteristics were included. Each of the 3 medical school characteristics significantly moderated the association between an applicant's CC attendance and the likelihood of application acceptance. The negative association between CC attendance and medical school application acceptance varies by medical school characteristics. Professional development for admissions officers focused on understanding the CC pathway and potential biases related to CC attendance would likely be beneficial in terms of trying to attract and select a diverse cohort of medical students in a holistic and equitable manner.

Measurement of a Large-Scale Short-Video Service Over Mobile and Wireless Networks

Short-video sharing services have seen explosive growth in recent years. Compared to conventional video sharing platforms, these have very different characteristics which are far from well-understood. This work aims at filling the gap by measuring and analyzing detailed application-level performance data from a top-10 short video service in China. The application-level data offered detailed and rare insights into many performance metrics of the service, which are otherwise inaccessible to external measurements. The service has a scale of over one billion daily views just for the mobile and wireless segments of the service. Our datasets covered over 22 billion video playbacks, over 100 million video files, served by over 5,000 servers to users across 35 provinces and 13 ISPs in China. We analyzed three aspects of the service: (a) video content characteristics; (b) network analytics; and (c) video streaming analytics. Our results revealed significant differences from conventional video-sharing platforms. These findings will have implications for system designs at all levels. The data also enabled us to conduct an indirect network performance measurement of mobile and wireless network services across China, as experienced by the service. These results offer rare insights into mobile and wireless networks' real-world performance in a large country.

Ambient Virtio: IO Virtualization for Seamless Integration and Access of Devices in Ambient Computing

Ambient computing creates a digital environment where a variety of technologies, including hardware, software, and even machine learning, are integrated into most useful and autonomous digital device that we want to use right here right now. The main obstacle for this computing is a barrier of vendor-specific details, programming interfaces, and access procedure even though all the things come to have network interfaces and interact with each other through the Internet. However, this configuration defies the expectation that ambient computing can reduce the demand for human attention. In this article, we propose <i>Ambient Virtio</i>, a device virtualization technology that virtualizes devices with the <small>virtio</small> framework. This allows devices in close proximity to users to be accessed without human intervention and integrated into metadevices that users want to create and access on the fly. We newly defined a <small>virtio</small>-based backend device named <small>virtio-ambient</small> and also constructed an ambient device architecture for virtual machines to read and write the remote device only with built-in system calls. We completely implemented the system and performed comparative evaluation with the existing application-level data networking system.

In-State Tuition Policies and the College Decisions of Undocumented Students: Evidence from Colorado

Abstract We study the effects of a decrease in college tuition on college application and enrollment behavior. Specifically, we use student-level data to analyze a Colorado law that granted in-state tuition to undocumented students residing in Colorado. We find an increase in the credit hours and persistence of newly enrolled and likely undocumented students. We do not find evidence of changes in the persistence or credit hours of continuing students. Leveraging application-level data, we show suggestive evidence that the policy induced more students to enroll in college through an increase in applications, rather than an increase in the acceptance rate or the enrollment rate.

Marketplace lending and consumer credit outcomes: evidence from prosper

ABSTRACT In 2005, Prosper launched the first peer-to-peer lending website in the US, allowing for consumers to apply for and receive loans entirely online. To understand the effect of this new credit source, we match application-level data from Prosper to credit bureau data. Post application, borrowers’ credit scores increase and their credit card utilization rates fall relative to non-borrowers in the short run. In the longer run, total debt levels for borrowers are higher than those of non-borrowers. Despite increased debt levels relative to non-borrowers, delinquency rates for borrowers are significantly lower.

Regularization-Based Coflow Scheduling in Optical Circuit Switches

To improve the application-level data efficiency, the scheduling of coflows, defined as a collection of parallel flows sharing the same objective, is prevailing in recent data centers. Meanwhile, optical circuit switches (OCS) are gradually applied to provide high data rate with low power consumption. However, so far few research outputs have covered the flow, let alone the coflow, scheduling in the context of OCS. In this work, we investigate coflow scheduling in OCS-based data centers. We first derive a novel operation called regularization processed respectively on the flow traffic demands and the flow start times, which can be efficiently implemented and reduce the circuit reconfiguration frequency dramatically. We then propose a 2-approximation algorithm, called Reco-Sin, for single coflow scheduling to minimize the coflow completion time (CCT). For multiple coflows, we derive Reco-Mul to minimize the total weighted CCT, which can transform any non-preemptive multi-coflow scheduling in packet switches to a scheduling scheme in OCS. Reco-Mul can achieve a constant approximation under the assumption that no tiny flows will be transmitted in OCS. To get rid of this assumption, we present another multiple coflow scheduling scheme, named Reco-Mul+, which has an approximation ratio of O(K). Here, K is the total number of coflows. Extensive simulations based on Facebook data traces show that our approaches outperform state-of-the-art schemes significantly, i.e., one single coflow can be finished up to 1.97× faster with Reco-Sin, and multiple coflows can be completed up to more than 2× faster with Reco-Mul and Reco-Mul+.

What does peer-to-peer lending evidence say about the Risk-Taking Channel of monetary policy?

This paper uses loan application-level data from a peer-to-peer lending platform to study the risk-taking channel of monetary policy. By employing a direct ex-ante measure of risk-taking and estimating the simultaneous equations of loan approval and loan amount, we provide evidence of monetary policy's impact on a nonbank financial institution's risk-taking. We find that the search-for-yield is the main driving force of the risk-taking effect, while we do not observe consistent findings of risk-shifting from the liquidity change. Monetary policy easing is associated with a higher probability of granting loans to risky borrowers and greater riskiness of credit allocation. However, these changes do not necessarily relate to a larger loan amount on average.

Fintech for the Poor: Financial Intermediation Without Discrimination*

Abstract I ask whether machine learning (ML) algorithms improve the efficiency in lending without compromising on equity in a credit environment where soft information dominates. I obtain loan application-level data from an Indian bank. To overcome the problem of the selective labels, I exploit the incentive-driven within officer difference in leniency within a calendar month. I find that the ML algorithm can lend 60% more at loan officers’ delinquency rate or achieve a 33% lower delinquency rate at loan officers’ approval rate. The efficiency is maintained even when the algorithm is explicitly prevented from discriminating against disadvantaged social classes.

Hotspot-Aware Hybrid Memory Management for In-Memory Key-Value Stores

Emerging Non-Volatile Memory (NVM) technologies promise much higher memory density and energy efficiency than DRAM, at the expense of higher read/write latency and limited write endurance. Hybrid memory systems composed of DRAM and NVM have the potential to provide very large capacity of main memory for in-memory key-value (K-V) stores. However, there remains challenges to directly deploy DRAM-based K-V stores in hybrid memory systems. The performance and energy efficiency of K-V stores on hybrid memory systems have not been fully explored yet. In this paper, we propose HMCached , an in-memory K-V store built on a hybrid DRAM/NVM system. HMCached utilizes an application-level data access counting mechanism to identify frequently-accessed (hotspot) objects (i.e., K-V pairs) in NVM, and migrates them to fast DRAM to reduce the costly NVM accesses. We also propose an NVM-friendly index structure to store the frequently-updated portion of object metadata in DRAM, and thus further mitigate the NVM accesses. Moreover, we propose a benefit-aware memory reassignment policy to address the slab calcification problem in slab-based K-V store systems, and significantly improve the benefit gain from the DRAM. We implement the proposed schemes with Memcached and evaluate it with Zipfian-like workloads. Experiment results show that HMCached significantly reduces NVM accesses by 70 percent compared to the vanilla Memcached running on a DRAM/NVM hybrid memory system without any optimizations, and improves application performance by up to 50 percent. Moreover, compared to a DRAM-only system, HMCached achieves 90 percent of performance and 46 percent reduction of energy consumption for realistic (read-intensive) workloads while significantly reducing the DRAM usage by 75 percent.

An Attack-Resilient Architecture for the Internet of Things

With current IoT architectures, once a single device in a network is compromised, it can be used to disrupt the behavior of other devices on the same network. Even though system administrators can secure critical devices in the network using best practices and state-of-the-art technology, a single vulnerable device can undermine the security of the entire network. The goal of this work is to limit the ability of an attacker to exploit a vulnerable device on an IoT network and fabricate deceitful messages to co-opt other devices. The approach is to limit attackers by using device proxies that are used to retransmit and control network communications. We present an architecture that prevents deceitful messages generated by compromised devices from affecting the rest of the network. The design assumes a centralized and trustworthy machine that can observe the behavior of all devices on the network. The central machine collects application layer data, as opposed to low-level network traffic, from each IoT device. The collected data is used to train models that capture the normal behavior of each individual IoT device. The normal behavioral data is then used to monitor the IoT devices and detect anomalous behavior. This paper reports on our experiments using both a binary classifier and a density-based clustering algorithm to model benign IoT device behavior with a realistic test-bed, designed to capture normal behavior in an IoT-monitored environment. Results from the IoT testbed show that both the classifier and the clustering algorithms are promising and encourage the use of application-level data for detecting compromised IoT devices.

Does Skillful Use of Hard Information by Machines Outperform a Combination of Hard and Soft Information of Loan Officers in Lending Decisions?

We ask whether machine learning algorithms improve the efficiency of screening of the loan officers, and thereby help expand access to formal credit. We obtain loan application level data from an Indian bank. To overcome the selective labels problem, we exploit the incentive driven within officer difference in leniency within a calendar month. We find that the ML algorithm is able to lend 26.6% more at loan officers' delinquency rate or achieve 21.3% lower delinquency rate at loan officers' approval rate. Higher efficiency is achieved without compromising on equity.

Marketplace Lending and Consumer Credit Outcomes: Evidence from Prosper

In 2005, Prosper launched the first peer-to-peer lending website in the US, allowing for consumers to apply for and receive loans entirely online. To understand the effect of this new credit source, we match application-level data from Prosper to credit bureau data. Post application, borrowers' credit scores increase and their credit card utilization rates fall relative to non-borrowers in the short run. In the longer run, total debt levels for borrowers are higher that of non-borrowers. Differences in mortgage debt are particularly large and increasing over time. Despite increased debt levels relative to non-borrowers, delinquency rates for borrowers are significantly lower.

The Examination of Continuation Applications and the Problem of Invalid Patents in the U.S.

It has long been recognised that the U.S. patent office routinely grants a large number of invalid patents. Recently, scholars have studied the root causes of this issue and have emphasised the key role played by the patent examination process itself. In this paper, I study how features of the examination process affect examiner behaviour and the problem of invalid patents in the context of continuation applications. These applications emanate from earlier patent applications filed at the patent office and allow applicants to submit new claims for a given invention. In the U.S., continuations are generally examined by the same examiner who was assigned to the earlier application. Using application-level data, I explore how this feature, which I call relatedness, affects examiners' grant decisions and other examination practices. I find that relatedness increases the grant rate, decreases examiners' efforts to narrow the scope of protection claimed by patent applicants and decreases the examiners' search efforts for prior art. I also show that the effects of relatedness do not seem to be driven by a wearing down effect on examiners and differ only slightly across different technology areas. Finally, I find evidence that relatedness leads to the granting of patents of more dubious validity. A key implication of these results is that the way the U.S. continuation application system is designed causes examiners to adopt softer examination practices, which in turn contributes to the problem of invalid patents.

A Data-Driven Algorithm for Indoor/Outdoor Detection Based on Connection Traces in a LTE Network

Environmental factors have a strong impact on the satisfaction of mobile users. Thus, estimating the context of a session is key to evaluating the end-user experience in mobile network management. Such a context is mainly defined by user location. In most cases, user location is derived from network measurements in the absence of handset measurements. Unfortunately, the current geolocation techniques do not have enough accuracy to detect if the user was indoor or outdoor. In this paper, a data-driven statistical model is proposed to detect if a cellular connection is originated in an indoor location based on the traffic attributes of the connection. Unlike the state-of-the-art approaches, based on application-level data, the proposed model is developed by logistic regression on data from radio connection traces stored in the network management system. The model is tested with a large trace dataset from a live Long Term Evolution (LTE) network.

What Does Peer-to-Peer Lending Evidence Say About the Risk-Taking Channel of Monetary Policy?

This paper uses loan application-level data from a peer-to-peer lending platform to study the risk-taking channel of monetary policy. By employing a direct ex-ante measure of risk-taking and estimating the simultaneous equations of loan approval and loan amount, we are the first to provide quantitative evidence of the impact of monetary policy on the risk-taking of nonbank financial institution. We find that the search-for-yield is the main workhorse of the risk-taking effect, while we do not observe consistent findings of risk-shifting from the liquidity change. Monetary policy easing is associated with a higher probability of granting loans to risky borrowers and a greater riskiness of credit allocation, but these changes do not necessarily relate to a larger loan amount on average.

What Does Peer-to-Peer Lending Evidence Say About the Risk-taking Channel of Monetary Policy?

This paper uses loan application-level data from a peer-to-peer lending platform to study the risk-taking channel of monetary policy. By employing a direct ex-ante measure of risk-taking and estimating the simultaneous equations of loan approval and loan amount, we are the first to provide quantitative evidence of the impact of monetary policy on the risk-taking of nonbank financial institution. We find that the search-for-yield is the main workhorse of the risk-taking effect, while we do not observe consistent findings of risk-shifting from the liquidity change. Monetary policy easing is associated with a higher probability of granting loans to risky borrowers and a greater riskiness of credit allocation, but these changes do not necessarily relate to a larger loan amount on average.

Knowledge Spillovers and Learning in the Workplace: Evidence from the U.S. Patent Office

Using application-level data from the Patent Office from 2001 to 2012, merged with personnel data on patent examiners, we explore the extent to which the key decision of examiners — whether to allow a patent — is shaped by the granting styles of her surrounding peers. Taking a number of methodological approaches to dealing with the common obstacles facing peer-effects investigations, we document strong evidence of peer influence. For instance, in the face of a one standard-deviation increase in the grant rate of her peer group, an examiner in her first two years at the Patent Office will experience a 0.15 standard-deviation increase in her own grant rate. Moreover, we document a number of markers suggesting that such influences arise, at least in part, through knowledge spillovers among examiners, as distinct from peer-pressure mechanisms. We even find evidence that some amount of these spillovers may reflect knowledge flows regarding specific pieces of prior art that bear on the patentability of the applications in question, as opposed to just knowledge flows regarding general examination styles. Finally, we find evidence suggesting that the magnitude of these peer examiner influences are just as strong, or stronger, than the influence of the examination styles of supervisors.

An energy-efficient system on a programmable chip platform for cloud applications

Performance analysis of master-slave based reconfigurable architecture and standalone SOPC based reconfigurable architecture using an analytical model.A massive-sessions optimized TCP/IP offload engine that supports up to 100K TCP sessions under 10Gbps line rate.An online dynamic scheduling method that can reconfigure or power off FPGA nodes according to workload variance to reduce the runtime energy consumption.Prototype of a reconfigurable cluster system based on standalone SOPC to provide cloud services, achieving up to 38X speed up in performance and 418X improvement in energy efficiency compared to the software based cloud systems. Traditional cloud service providers build large data-centers with a huge number of connected commodity computers to meet the ever-growing demand on performance. However, the growth potential of these data-centers is limited by their corresponding energy consumption and thermal issues. Energy efficiency becomes a key issue of building large-scale cloud computing centers. To solve this issue, we propose a standalone SOPC (System on a Programmable Chip) based platform for cloud applications. We improve the energy efficiency for cloud computing platforms with two techniques. First, we propose a massive-sessions optimized TCP/IP hardware stack using a macro-pipeline architecture. It enables the hardware acceleration of pipelining execution of network packet offloading and application level data processing. This achieves higher energy efficiency while maintaining peak performance. Second, we propose a online dynamic scheduling strategy. It can reconfigure or shut down FPGA nodes according to workload variance to reduce the runtime energy consumption in a standalone SOPC based reconfigurable cluster system. Two case studies including a webserver application and a cloud based ECG (electrocardiogram) classification application are developed to validate the effectiveness of the proposed platform. Evaluation results show that our SOPC based cloud computing platform can achieve up to 418X improvement in terms of energy efficiency over commercial cloud systems.

Energy efficient architecture for graph analytics accelerators

Specialized hardware accelerators can significantly improve the performance and power efficiency of compute systems. In this paper, we focus on hardware accelerators for graph analytics applications and propose a configurable architecture template that is specifically optimized for iterative vertex-centric graph applications with irregular access patterns and asymmetric convergence. The proposed architecture addresses the limitations of the existing multi-core CPU and GPU architectures for these types of applications. The SystemC-based template we provide can be customized easily for different vertex-centric applications by inserting application-level data structures and functions. After that, a cycle-accurate simulator and RTL can be generated to model the target hardware accelerators. In our experiments, we study several graph-parallel applications, and show that the hardware accelerators generated by our template can outperform a 24 core high end server CPU system by up to 3x in terms of performance. We also estimate the area requirement and power consumption of these hardware accelerators through physical-aware logic synthesis, and show up to 65x better power consumption with significantly smaller area.

Dealing with Data Quality in Smart Home Environments—Lessons Learned from a Smart Grid Pilot

Over the last years, we have witnessed increasing interconnection between the physical and digital world. The so called Internet of Things (IoT) is becoming more and more a reality in application domains like manufacturing, mobile computing, transportation, and many others. However, despite promising huge potential, the application domain of smart homes is still at its infancy and lags behind other fields of IoT. A deeper understanding of this type of techno-human system is required to make this vision a reality. In this paper, we report findings from a three year pilot that sheds light on the challenges of leveraging IoT technology in the home environment. In particular, we provide details on data quality issues in real-world deployments. That is, we analyze application level data for errors in measurements as well as issues in the end-to-end communication. Understanding what data errors to expect is crucial for understanding the smart building domain and paramount for building successful applications. With our work, we provide insights in a domain of IoT that has tremendous growth potential and help researchers as well as practitioners to better account for the data characteristics of smart homes.