Predictive Models In Order Research Articles

Development of a depression in Parkinson's disease prediction model using machine learning.

BACKGROUNDIt is important to diagnose depression in Parkinson’s disease (DPD) as soon as possible and identify the predictors of depression to improve quality of life in Parkinson’s disease (PD) patients.AIMTo develop a model for predicting DPD based on the support vector machine, while considering sociodemographic factors, health habits, Parkinson's symptoms, sleep behavior disorders, and neuropsychiatric indicators as predictors and provide baseline data for identifying DPD.METHODSThis study analyzed 223 of 335 patients who were 60 years or older with PD. Depression was measured using the 30 items of the Geriatric Depression Scale, and the explanatory variables included PD-related motor signs, rapid eye movement sleep behavior disorders, and neuropsychological tests. The support vector machine was used to develop a DPD prediction model.RESULTSWhen the effects of PD motor symptoms were compared using “functional weight”, late motor complications (occurrence of levodopa-induced dyskinesia) were the most influential risk factors for Parkinson's symptoms. CONCLUSIONIt is necessary to develop customized screening tests that can detect DPD in the early stage and continuously monitor high-risk groups based on the factors related to DPD derived from this predictive model in order to maintain the emotional health of PD patients.

Electrocoagulation Based Chromium Removal Efficiency Classification Using Logistic Regression

Surface treatment and tanning industries use huge quantities of heavy metals—especially Chromium (III) and (VI)—in their processes thanks to its physical proprieties. It is used in the composition of special steels and refractory alloys. By dint of using this metal, an enormous quantity of rejects is produced each year and discharged into the oceans. As this is very dangerous for our environment, it is very important to treat these discharges before getting rid of them. This study treats chromium removal as a special type of heavy metals that can be a component of industrial discharges. Electrocoagulation is considered among the best methods used in this kind of treatment. However, it requires a lot of time, energy and remains expensive. This paper presents a predictive model in order to classify the chromium removal efficiency using electrocoagulation method. The proposed model is a logistic regression (LR) that consumes four parameters that we call predictors: pH, time, current, and stirring speed. After the training and validation process, we obtained 88% as classification precision, recall and F-Score metrics values while the use of the 10-Folds cross-validation method gave a minimal area under curve (AUC) value of 97% while the best value attempts 100%. Classification report states that the model performs well comparing to similar experimentation efficiencies.

Experimental investigation of structural steel beams subjected to localized fire

This paper presents an experimental behavior of unprotected W16 × 26 steel beams subjected to localized fire. Beam specimens were either simply supported or connected to steel columns via all bolted double angle connections. Two test fires, steady or transient heating, were produced using a natural gas fueled burner located below the center of the beams. Thermal and structural responses of the steel beams as well as the heat release rate and applied forces were measured. The tests showed that all specimens failed by lateral-torsional buckling regardless of fire loading and end support conditions. The results highlight that the heating conditions and end restraints influence the behavior of steel beams under localized fire. The experimental results presented can be used for validation of predictive models in order to improve deign methods for localized fire conditions.

Novel clustering-based pruning algorithms

One of the crucial problems of designing a classifier ensemble is the proper choice of the base classifier line-up. Basically, such an ensemble is formed on the basis of individual classifiers, which are trained in such a way to ensure their high diversity or they are chosen on the basis of pruning which reduces the number of predictive models in order to improve efficiency and predictive performance of the ensemble. This work is focusing on clustering-based ensemble pruning, which looks for the group of similar classifiers which are replaced by their representatives. We propose a novel pruning criterion based on well-known diversity measures and describe three algorithms using classifier clustering. The first method selects the model with the best predictive performance from each cluster to form the final ensemble, the second one employs the multistage organization, where instead of removing the classifiers from the ensemble each classifier cluster makes the decision independently, while the third proposition combines multistage organization and sampling with replacement. The proposed approaches were evaluated using 30 datasets with different characteristics. Experimentation results validated through statistical tests confirmed the usefulness of the proposed approaches.

SPELEOLOGICAL POTENTIAL OF THE SUMIDOURO STATE PARK / MG

Thematic mapping for the speleological potential of the Sumidouro State Park / mg. The Sumidouro State Park (PESU) constitutes a Conservation Unit located in the Environmental Protection Area (APA) Karst Lagoa Santa, between the municipalities of Lagoa Santa and Pedro Leopoldo, northern portion of the Metropolitan Region of Belo Horizonte, housing a rich speleological heritage. Currently, there are 49 known cavities registered in the database of the National Center for Research and Conservation of Caves (CECAV). It is estimated that there are more caves inserted in the park not yet registered by the agency. Thus, this work aimed to develop a predictive model in order to indicate areas with a higher potential for the occurrence of caves, in the light of geotechnologies. In order to perform the speleological potential calculation, data were collected on the existing features that are intrinsically linked to the formation of caves: geology, hydrography, exokarstic features present in the study area, hypsometry and slope. After data acquisition, values were assigned to them and, through map algebra, it was possible to identify areas with a higher and very higher degree of potential for the occurrence of caves. Because of the development of this model, it is expected that it will have application in other areas of study. Future fieldwork should be carried out to confirm and register new cavities pointed out in the mapped areas, especially those registered as having high potential. Keywords: karst; Apa; Lagoa Santa; speleological potential; multicriteria analysis.

Combining predictive base models using deep ensemble learning

Since information science and communication technologies had improved significantly, data volumes had expanded. As a result of that situation, advanced pre-processing and analysis of collected data became a crucial topic for extracting meaningful patterns hidden in the data. Therefore, traditional machine learning algorithms generally fail to gather satisfactory results when analyzing complex data. The main reason of this situation is the difficulty of capturing multiple characteristics of the high dimensional data. Within this scope, ensemble learning enables the integration of diversified single models to produce weak predictive results. The final combination is generally achieved by various voting schemes. On the other hand, if a large amount of single models are utilized, voting mechanism cannot be able to combine these results. At this point, Deep Learning (DL) provides the combination of the ensemble results in a considerable time. Apart from previous studies, we determine various predictive models in order to forecast the outcome of two different case studies. Consequently, data cleaning and feature selection are conducted in advance and three predictive models are defined to be combined. DL based integration is applied substituted for voting mechanism. The weak predictive results are fused based on Recurrent Neural Network (RNN) and Long Short Term Memory (LSTM) using different parameters and datasets and best predictors are extracted. After that, different experimental combinations are evaluated for gathering better prediction results. For comparison, grouped individual results (clusters) with proper parameters are compared with DL based ensemble results.

Multimodal neuroimaging-based prediction of adult outcomes in childhood-onset ADHD using ensemble learning techniques.

Attention-deficit/hyperactivity disorder (ADHD) is a highly prevalent and heterogeneous neurodevelopmental disorder, which is diagnosed using subjective symptom reports. Machine learning classifiers have been utilized to assist in the development of neuroimaging-based biomarkers for objective diagnosis of ADHD. However, existing basic model-based studies in ADHD report suboptimal classification performances and inconclusive results, mainly due to the limited flexibility for each type of basic classifier to appropriately handle multi-dimensional source features with varying properties. This study applied ensemble learning techniques (ELTs), a meta-algorithm that combine several basic machine learning models into one predictive model in order to decrease variance, bias, or improve predictions, in multimodal neuroimaging data collected from 72 young adults, including 36 probands (18 remitters and 18 persisters of childhood ADHD) and 36 group-matched controls. All currently available optimization strategies for ELTs (i.e., voting, bagging, boosting and stacking techniques) were tested in a pool of semifinal classification results generated by seven basic classifiers. The high-dimensional neuroimaging features for classification included regional cortical gray matter (GM) thickness and surface area, GM volume of subcortical structures, volume and fractional anisotropy of major white matter fiber tracts, pair-wise regional connectivity and global/nodal topological properties of the functional brain network for cue-evoked attention process. As a result, the bagging-based ELT with the base model of support vector machine achieved the best results, with significant improvement of the area under the receiver of operating characteristic curve (0.89 for ADHD vs. controls and 0.9 for ADHD persisters vs. remitters). Features of nodal efficiency in right inferior frontal gyrus, right middle frontal (MFG)-inferior parietal (IPL) functional connectivity, and right amygdala volume significantly contributed to accurate discrimination between ADHD probands and controls; higher nodal efficiency of right MFG greatly contributed to inattentive and hyperactive/impulsive symptom remission, while higher right MFG-IPL functional connectivity strongly linked to symptom persistence in adults with childhood ADHD. Considering their improved robustness than the commonly implemented basic classifiers, findings suggest that ELTs may have the potential to identify more reliable neurobiological markers for neurodevelopmental disorders.

Susceptible, Infectious and Recovered (SIR Model) Predictive Model to Understand the Key Factors of COVID-19 Transmission

On 31 December 2019, WHO was alerted to several cases of pneumonia in Wuhan City, Hubei Province of China. The virus did not match any other known virus. This raised concern because when a virus is new, general behavior and how it affects, people do not know. Initial few cases reportedly had some link to a large seafood and animal market, suggesting animal-to-person spread. However, a growing number of patients reportedly have not had exposure to animal markets, indicating person-to-person spread is occurring. At this time, it’s unclear how easily or sustainably this virus is spreading between people. At any given time during a flu epidemic, firstly, should know the number of people who are infected. Second, to know the numbers who have been infected and have recovered, because these people now have immunity to the disease. Well established SIR modeling methodology is used to develop a predictive model in order to understand the key factors that impact the COVID-19 transmission.

Temporal analysis for dropout prediction using self-regulated learning strategies in self-paced MOOCs

MOOCs (Massive Open Online Courses) have usually high dropout rates. Many articles have proposed predictive models in order to early detect learners at risk to alleviate this issue. Nevertheless, existing models do not consider complex high-level variables, such as self-regulated learning (SRL) strategies, which can have an important effect on learners' success. In addition, predictions are often carried out in instructor-paced MOOCs, where contents are released gradually, but not in self-paced MOOCs, where all materials are available from the beginning and users can enroll at any time. For self-paced MOOCs, existing predictive models are limited in the way they deal with the flexibility offered by the course start date, which is learner dependent. Therefore, they need to be adapted so as to predict with little information short after each learner starts engaging with the MOOC. To solve these issues, this paper contributes with the study of how SRL strategies could be included in predictive models for self-paced MOOCs. Particularly, self-reported and event-based SRL strategies are evaluated and compared to measure their effect for dropout prediction. Also, the paper contributes with a new methodology to analyze self-paced MOOCs when carrying out a temporal analysis to discover how early prediction models can serve to detect learners at risk. Results of this article show that event-based SRL strategies show a very high predictive power, although variables related to learners' interactions with exercises are still the best predictors. That is, event-based SRL strategies can be useful to predict if e.g., variables related to learners' interactions with exercises are not available. Furthermore, results show that this methodology serves to achieve early powerful predictions from about 25 to 33% of the theoretical course duration. The proposed methodology presents a new approach to predict dropouts in self-paced MOOCs, considering complex variables that go beyond the classic trace-data directly captured by the MOOC platforms.

Exploring crime patterns in Mexico City

IntroductionLife in the city generates data on human behavior in many different ways. Measuring human behavior in terms of criminal offenses plays a critical role on identifying the most common crime type in each urban area. A primary concern for the population of the largest cities in the world is to avoid specific city zones that could show a significant risk. This case of study systematically explores different sources of data including social media related to crime reports. In addition, this research seeks to examine and predict the most frequent crimes in Mexico City.Case descriptionThis case study was conducted in the form of an exploratory research. One of the parties involved is the Mexico City Police Department which released the approximate locations and categories of all crime reports from January 2013 to September 2016. We analyze the impact of crime in Mexico City based on 13 official crime categories: Robbery passerby, Theft of motor vehicle, Robbery of business property, Card fraud, Homicide, Domestic burglary, Robbery on public transportation, Rape, Firearm injuries, Robbery in subway, Robbery on taxi, Robbery to carrier, and Robbery to deliver person. We compare and analyze how people report a crime through the traditional system and using social media.Discussion and evaluationThis research uses a quantitative case study approach to investigate, how our predictive model is able to forecast the total number of reported crimes in the following week based on its previous weekly aggregated observations and Google Trends series. Similarly, this case study seeks to determine whether Twitter performs correctly as a “social crime sensor” in terms of detecting certain areas and boroughs that are more likely to show criminal behavior.ConclusionsIn this study we used a linear predictive model in order to evaluate the performance of Google Trends in predicting crime rates based on weekly analysis. In addition, Twitter showed a suitable performance to discover the spatial distribution of crime frequency in Mexico City. Finally, this study provides an important opportunity to develop and encourage tailored strategies to tackle crime.

Predicting the Impact of Climate Change on Thermal Comfort in A Building Category: The Case of Linear-type Social Housing Stock in Southern Spain

The Climate Change scenario projected by the IPCC for the year 2050 predicts noticeable increases in temperature. In severe summer climates, such as the Mediterranean area, this would have very negative effects on thermal comfort in the existing housing stock, given the current high percentage of dwellings which are obsolete in energy terms and house a population at serious risk of energy poverty. The main aim of this paper is to generate a predictive model in order to assess the impact of this future climate scenario on thermal comfort conditions in an entire building category. To do so, calibrated models representing linear-type social multi-family buildings, dating from the post-war period and located in southern Spain, will be simulated extensively using transient energy analyses performed by EnergyPlus. In addition, a sensitivity analysis will be performed to identify the most influential parameters on thermal discomfort. The main results predict a generalized deterioration in indoor thermal comfort conditions due to global warming, increasing the average percentage of discomfort hours during the summer by more than 35%. This characterization of the future thermal behaviour of the residential stock in southern Spain could be a trustworthy tool for decision-making in energy retrofitting projects which are so badly needed. To do so, further work is required on some limitations of this model so that different user profiles and typologies can be represented in detail and an economic assessment can be included.

Predicting psychogenic non-epileptic seizures from serum levels of neuropeptide Y and adrenocorticotropic hormone.

Patients with psychogenic non-epileptic seizures (PNES) may present with convulsive events that are not accompanied by epileptiform brain activity. Video-electroencephalography (EEG) monitoring is the gold standard for diagnosis, yet not all patients experience convulsive episodes during video-EEG sessions. Hence, we aimed to construct a predictive model in order to detect PNES from serum hormone levels, detached from an evaluation of patients' convulsive episodes. Fifteen female patients with PNES and 60 healthy female controls participated in the study, providing blood samples for hormone analysis. A binomial logistic regression model and the leave-one-out cross-validation were employed. We found that levels of neuropeptide Y and adrenocorticotropic hormone were the optimal combination of predictors, with over 90% accuracy (area under the curve=0.980). The ability to diagnose PNES irrespective of convulsive events would represent an important step considering its feasibility and affordability in daily clinical practice.

Single-Use System Integrity I: Using a Microbial Ingress Test Method to Determine the Maximum Allowable Leakage Limit (MALL).

An aerosol microbial ingress test was specifically designed and used to create a predictive model in order to determine the maximum allowable leakage limit (MALL) of single-use systems (SUSs). The MALL is defined as the greatest leak size that does not pose any risk to the product. The procedure involved taking test samples of film material from single-use bags. As test samples, an ethylene vinyl acetate multilayer film (300 μm thick) and a polyethylene multilayer film (400 μm thick) were cut into 50 mm patches. Artificial defects of 1-100 µm were laser-drilled in the middle of each film patch. The patch was assembled on a holder and properly sealed. The test units were filled aseptically with culture media and placed inside an aerosol chamber. Various pressures were applied to the test unit to simulate the constraints that single-use systems may be subject to under real-world conditions. After an aerosolization cycle with spores of Bacillus atrophaeus, a minimum concentration of 106 CFU/cm2 was reached on the film surface. The test units were incubated for 14 days at 30°C-35°C and visually inspected for bacterial ingress. Thirty samples per defect size were tested. Logistic regression was used to indicate the MALL for a single-use system according to the required risk level. With this method, the probability of the occurrence or absence of ingress for a specific defect size was reported according to the experimental data. In addition to physical parameters, such as the pressure applied and the film material, the effect of the probabilistic nature of the microorganisms in determining the MALL is considered. Although finding an experimental model to predict the MALL for real-life process conditions was the ultimate objective, this paper also presents the microbial ingress test data obtained so far for two extreme conditions. Potential constraints, such as vibration, shock, acceleration, liquid movement, and pressure differentials, observed during normal usage were simulated using two extreme differential pressures, 0 mbar and 300 mbar. The estimated MALL for typical use-case conditions are 10-20 µm for storage applications and 2-10 µm for shipping conditions. The microbial integrity test method used in this article was able to detect bacterial ingress down to 3 µm defect size.LAY ABSTRACT: As use of single-use systems (SUSs) is increasingly expanding into all process steps of commercial manufacturing, integrity failure can significantly impact drug safety, availability, and costs. Consequently, growing industry scrutiny on single-use system integrity (SUSI) is raising the need to develop good science behind reliable determinations of liquid leakage and microbial ingress as well as the appropriate physical integrity testing technologies. In the current study, microbial ingress testing by the aerosol method is used to determine the maximum allowable leakage limit (MALL) for SUSs. To define the MALL, it is generally assumed that a system or product will not show any microbial ingress or leakage at a certain defect size. Statistical analysis of the experimental data in this study indicated the MALL with probability at a certain defect size for each system. As a result, the method studied provides a more accurate way of predicting ingress and increasing safety down the line for drug manufacturers and patients alike.

Prognostic Impact of Presurgical CA19-9 Level in Pancreatic Adenocarcinoma: A Pooled Analysis

BACKGROUND: Presurgical carbohydrate antigen 19-9 (CA19-9) level predicts overall survival (OS) in resected pancreatic adenocarcinoma (PaC). The aim of this pooled analysis was to evaluate if presurgical CA19-9 level can also predict local control (LC) and distant metastasis-free survival (DMFS). METHODS: Seven hundred patients with PaC from eight institutions who underwent surgical resection ± adjuvant treatment between 2000 and 2014 were analyzed. Patients were divided based on four presurgical CA19-9 level cutoffs (5, 37, 100, 353 U/ml). Weibull regression model to identify independent predictors of OS on 404 patients with complete information was fitted. RESULTS: Median follow-up was 17 months (range: 2-225 months). Univariate analysis showed a better prognosis in pT1-2, pN0, diameter <30 mm, or grade 1 tumors and in patients undergoing R0 resection, distal pancreatectomy, or adjuvant chemotherapy and with lower CA19-9 levels. Five-year OS, LC, and DMFS were as follows: CA19-9 <5.0: 5.7%, 47.2%, 17.0%; CA19-9 5.1-37.0: 37.9%, 63.3%, 46.0%; CA19-9 37.1-100.0: 27.1%, 59.4%, 39.0%; CA19-9 100.1-353.0: 17.4%, 43.4%, 26.7%; CA19-9 >353.1: 10.9%, 50.2%, and 23.4%, respectively. At multivariate analysis, CA19-9 >100 and <353 level (P=.002), CA19-9 ≥353.1 (P<.001) level, G3 tumor (P=.002), and tumor diameter >30 mm (P<.001) correlated with worse OS. Patients treated with postoperative chemoradiation doses >50.0 Gy showed improved OS (P<.001). CONCLUSION: Presurgical CA19-9 predicts both OS and pattern of failure. Therefore, CA19-9 should be included in predictive models in order to customize treatments based on prognostic factors. Moreover, future studies should stratify patients according to presurgical CA19-9 level.

Prediction of central lymph node metastasis in papillary thyroid microcarcinoma according to clinicopathologic factors and thyroid nodule sonographic features: a case-control study.

BackgroundPreoperative diagnosis of central lymph node metastasis (CLNM) poses to be a challenge in clinical node-negative papillary thyroid microcarcinoma (PTMC). This research work aims at investigating the association existing between BRAF mutation, clinicopathological factors, ultrasound characteristics, and CLNM, in addition to establishing a predictive model for CLNM in PTMC.Materials and methodsThe study included 673 PTMC patients, already undergone total thyroidectomy or lobectomy with prophylactic central lymph node dissection. The predictor factors were identified through univariate and multivariate analyses. The support vector machine was put to use to develop statistical models, which could predict CLNM on the basis of independent predictors.ResultsTumor size (>5 mm), lower location, no well-defined margin, contact of >25% with the adjacent capsule, display of enlarged lymph nodes, and BRAF mutation were independent predictors of CLNM. Through the use of the predictive model, 79.6% of the patients were classified accurately, the sensitivity and specificity amounted to be 85.1% and 75.8%, respectively, and the positive predictive value and negative predictive value stood at 71.6% and 87.6%, respectively.ConclusionsWe established a predictive model in order to predict CLNM preoperatively in PTMC when preoperative diagnosis of CLNM was not clear.

Prediction of Topsoil Texture Through Regression Trees and Multiple Linear Regressions

Users of soil survey products are mostly interested in understanding how soil properties vary in space and time. The aim of digital soil mapping (DSM) is to represent the spatial variability of soil properties quantitatively to support decision-making. The goal of this study is to evaluate DSM techniques (Regression Trees - RT and Multiple Linear Regressions - MLR) and the ability of these tools to predict mineral fraction content under a wide variability of landscapes. The study site was the entire Guapi-Macacu watershed (1,250.78 km²) in the state of Rio de Janeiro in the Southeast region of Brazil. Terrain attributes and remote sensing data (with 30 m of spatial resolution) were used to represent landscape co-variables selected as an input in predictive models in order to develop the explanatory variables. The selection of sampling sites was based on the Latin Hypercube algorithm. A representative set of one hundred points with feasible field access was chosen. Different input databases were tested for prediction of mineral fraction content (harmonized and original data). The Spline algorithm was used to harmonize data according to the GlobalSoil.Net consortium standards. The results showed better performance from the RT models, using input from an average of six covariates; the simplest MLR model used twice as many input variables, creating more complex models without gaining precision. Furthermore, better R² values were obtained using RT models, irrespective of harmonization of soil data. The harmonized dataset from the 0.00-0.05 and 0.05-0.15 m layers, in general, presented better results for the clay and silt, with R2 values of 0.52 (0.00-0.05 m) and 0.69 (0.05-0.15 m), respectively. Prediction of sand content showed better results when the original depth data was used as an input, although all regression tree models had R2 values greater than 0.52. The RT models provided a better statistical index than MLR for all predicted properties; however, the variance between models suggests similarity of performance. Regarding harmonization of soil data, both input databases (harmonized or not) can be used to predict soil properties, since the variance of model performance was low and generalization of the soil maps showed similar trends. The products obtained from the digital soil mapping approach make it possible to integrate the factor of uncertainties, providing easier interpretation for soil management and land use decisions.

Analysis of Construction Auctions Data in Slovak Public Procurement

The requirement for efficient public spending leads contracting authorities to use electronic reverse auctions (e‐RA), a tool that allows achieving financial savings. In this study, we aim to explore the relationships between the different acting e‐RA variables and to check for predictive models in order to infer on the savings amount in construction public procurement. Data on real construction auctions in Slovakia were statistically analysed by means of graphics tools, multiple regression analysis, test, and statistics for measuring the association between categorical variables. The results revealed that one should take the type of contract into account when considering the use of e‐RA. This research provides several implications for purchasing practitioners in the area of construction procurement, especially with regard to the level of competition in the auction and estimation of savings potential. Presented findings aid managerial decision‐making process of e‐RA adoption. At the end, recommended future research directions in the investigated area are outlined.

Prediction of GTV median dose differences eases Monte Carlo re-prescription in lung SBRT

Background and PurposeThe use of Monte Carlo (MC) dose calculation algorithm for lung patients treated with stereotactic body radiotherapy (SBRT) can be challenging. Prescription in low density media and time-consuming optimization conducted CyberKnife centers to propose an equivalent path length (EPL)-to-MC re-prescription method based on GTV median dose. Unknown at the time of planning, GTV D50% practical application remains difficult. The current study aims at creating a re-prescription predictive model in order to limit conflicting dose value during EPL optimization. Material and Methods129 patients planned with EPL algorithm were recalculated with MC. Relative GTV_D50% discrepancies were assessed and influencing parameters identified using wrapper feature selection. Based on best descriptive parameters, predictive nomogram was built from multivariate linear regression. EPL-to-MC OARs near max-dose discrepancies were reported. ResultsThe differences in GTV_D50% (median 10%, SD: 9%) between MC and EPL were significantly (p < .001) impacted by the lesion’s surface-to-volume ratio and the average relative electronic density of the GTV and the GTV’s 15 mm shell. Built upon those parameters, a nomogram (R2 = 0.79, SE = 4%) predicting the GTV_D50% discrepancies was created. Furthermore EPL-to-MC OAR dose tolerance limit showed a strong linear correlation with coefficient range [0.84–0.99]. ConclusionGood prediction on the required re-prescription can be achieved prior planning using our nomogram. Based on strong linear correlation between EPL and MC for OARs near max-dose, further restriction on dose constraints during the EPL optimization can be warranted. This a priori knowledge eases the re-prescription process in limiting conflicting dose value.

Estimating indoor galaxolide concentrations using predictive models based on objective assessments and data about dwelling characteristics

Background: Galaxolide (HHCB) is used for fragrance in many consumer products. The aim of the current study was to use objective assessments of HHCB to build a predictive model in order to estimate indoor-measured HHCB concentrations from questionnaire-based data on dwelling characteristics and occupants’ habits and activities.Methods: Environmental assessments of indoor HHCB, dwelling characteristics were carried out in 150 dwellings in Brittany (France). Among the various models that were tested, the best predictive model for the reduced set of characteristics was identified on the basis of the coefficient of determination (R2) criterion.Results: Linear regression model showed among the best performances (R2 = 0.48), together with some more complex models. According to the estimated results, the main variables that significantly increased HHCB concentrations were: living in rural area, drying clothes inside dwellings, painted walls, chipboard furniture, double glazing, damaged floors and duration of bathroom door being kept open. Laminated floors and presence of indoor plants were found to significantly decrease HHCB concentrations.Discussion: The linear model based on objective assessments and questionnaire-derived data about dwelling characteristics and occupants’ activities constituted an easy method for predicting indoor air HHCB concentrations. For studies including a large number of dwellings, modeling of HHCB concentrations is cheaper than measuring it in every location. Our methodological procedure can be applied to other indoor air pollutants.

Algorithmic Transparency for the Smart City

Emerging across many disciplines are questions about algorithmic ethics – about the values embedded in artificial intelligence and big data analytics that increasingly replace human decisionmaking. Many are concerned that an algorithmic society is too opaque to be accountable for its behavior. An individual can be denied parole or denied credit, fired or not hired for reasons she will never know and cannot be articulated. In the public sector, the opacity of algorithmic decisionmaking is particularly problematic both because governmental decisions may be especially weighty, and because democratically-elected governments bear special duties of accountability. Investigative journalists have recently exposed the dangerous impenetrability of algorithmic processes used in the criminal justice field – dangerous because the predictions they make can be both erroneous and unfair, with none the wiser. We set out to test the limits of transparency around governmental deployment of big data analytics, focusing our investigation on local and state government use of predictive algorithms. It is here, in local government, that algorithmically-determined decisions can be most directly impactful. And it is here that stretched agencies are most likely to hand over the analytics to private vendors, which may make design and policy choices out of the sight of the client agencies, the public, or both. To see just how impenetrable the resulting “black box” algorithms are, we filed 42 open records requests in 23 states seeking essential information about six predictive algorithm programs. We selected the most widely-used and well-reviewed programs, including those developed by for-profit companies, nonprofits, and academic/private sector partnerships. The goal was to see if, using the open records process, we could discover what policy judgments these algorithms embody, and could evaluate their utility and fairness. To do this work, we identified what meaningful “algorithmic transparency” entails. We found that in almost every case, it wasn’t provided. Over-broad assertions of trade secrecy were a problem. But contrary to conventional wisdom, they were not the biggest obstacle. It will not usually be necessary to release the code used to execute predictive models in order to dramatically increase transparency. We conclude that publicly-deployed algorithms will be sufficiently transparent only if (1) governments generate appropriate records about their objectives for algorithmic processes and subsequent implementation and validation; (2) government contractors reveal to the public agency sufficient information about how they developed the algorithm; and (3) public agencies and courts treat trade secrecy claims as the limited exception to public disclosure that the law requires. Although it would require a multi-stakeholder process to develop best practices for record generation and disclosure, we present what we believe are eight principal types of information that such records should ideally contain.