Random Forrest Research Articles

Predicting preterm birth using auto-ML frameworks: a large observational study using electronic inpatient discharge data.

To develop and compare different AutoML frameworks and machine learning models to predict premature birth. The study used a large electronic medical record database to include 715,962 participants who had the principal diagnosis code of childbirth. Three Automatic Machine Learning (AutoML) were used to construct machine learning models including tree-based models, ensembled models, and deep neural networks on the training sample (N = 536,971). The area under the curve (AUC) and training times were used to assess the performance of the prediction models, and feature importance was computed via permutation-shuffling. The H2O AutoML framework had the highest median AUC of 0.846, followed by AutoGluon (median AUC: 0.840) and Auto-sklearn (median AUC: 0.820), and the median training time was the lowest for H2O AutoML (0.14 min), followed by AutoGluon (0.16 min) and Auto-sklearn (4.33 min). Among different types of machine learning models, the Gradient Boosting Machines (GBM) or Extreme Gradient Boosting (XGBoost), stacked ensemble, and random forrest models had better predictive performance, with median AUC scores being 0.846, 0.846, and 0.842, respectively. Important features related to preterm birth included premature rupture of membrane (PROM), incompetent cervix, occupation, and preeclampsia. Our study highlights the potential of machine learning models in predicting the risk of preterm birth using readily available electronic medical record data, which have significant implications for improving prenatal care and outcomes.

Automated spinal MRI-based diagnostics of disc bulge and desiccating using LS-RBRP with RF

<p>Low back pain occurs because of the degeneration in Intervertebral Disc (IVD) namely: Disc Desiccation, Disc Bulge, and Disc Herniation, etc. To detect disc degeneration, a doctor often physically evaluates the Magnetic Resonance Imaging (MRI), which takes time and is dependent on the doctor’s expertise and training. Degeneration diagnosis that is automated can ease some of the doctor’s workload. On 378 IVDs for 63 patients, the proposed method is trained, tested, and assessed. According to the performance evaluation, the proposed Local Sub-Rhombus Binary Relationship (LS-RBRP) and Random Forrest (RF) classifier approach gives an overall accuracy of 90.2%. The proposed approach also produces a higher sensitivity, specificity, precision, and F-score of 80.8%, 90.3%, 90.4%, and 84.5%, respectively, when diagnosing the normal IVD, disc desiccation, and disc bulge in the lumbar MRI.</p>

#3974 EFFECTIVE PREDICTION OF NEED FOR CHRONIC DIALYSIS IN CHILDREN, BASED ON RANDOM FORREST ALGORITHMS USING BLOOD CELL COUNT-DRIVEN PARAMETERS

Abstract Background and Aims Persistent, low-grade inflammation is a significant component of chronic kidney disease (CKD) that plays a pivotal role in its pathophysiology, progression, complications, and all-cause mortality. Previous studies have searched for simple, cost-effective, and universally available markers useful in the systemic inflammation assessment in CKD patients. In the adult population, an increase in neutrophil count, correlated with a reduction in lymphocyte count, indicates the rate of progression to dialysis and predicts mortality in hemodialysis and peritoneal dialysis patients. Due to modulating role of platelets within the inflammatory pathways, a mean platelet volume (MPV) and platelet-to-lymphocyte ratio have also been proposed as markers of inflammation. The systemic immune inflammation index (SII) is a newly defined ratio combining neutrophil, lymphocyte, and platelet counts. It is proposed as a prognostic indicator comprehensively reflecting patients' inflammatory and immune status. Recently, elevated SII has shown a predictive value for mortality risk among CKD adult patients. SII has not been evaluated in the pediatric CKD population so far. Thus, the study aimed to analyze complete blood cell count (CBC) driven parameters, including SII, in children with CKD and to assess their potential usefulness in the prediction of the need for chronic replacement therapy with the use of artificial intelligence tools. Method The study group consisted of 27 predialysis children with CKD stages 4-5 (stage 4 - 11 patients, stage 15 - 16 children) and 40 patients on chronic dialysis (HD - 21 children, APD - 19 patients). The patients’ age ranged from 5 to 18 years, children under 5 were excluded due to the different CBC profile regarding neutrophil – lymphocyte proportions. The evaluated CBC parameters were: hemoglobin, hematocrit, leukocyte, neutrophil, monocyte, lymphocyte, platelet counts and MPV. Kidney function, standard biochemical parameters, CRP and SII were also analyzed. This database was used for further analysis. The recursively selected subsets of input variables constituted the input of random forrest classifier (RFC). Each variant has been optimized, taking into account various features (accuracy, AUROC, precision, recall, MCC), then the best model was saved. The aim was to narrow the set of input parameters so that high predictive power is secured, whereas overfitting or overcomplicating could be avoided. Moreover, the GINI importance was measured in order to define the parameter with the largest share in the prediction. Results The best Random Forest Classifier contained neutrophil count, MPV, and SII as input variables, and achieved the following values: AUROC 0.9286, accuracy 93.75%, precision 0.9437, recall 0.9375 and MCC 0.87. The statistics for each class were as follows: precision 0.90, recall 1.00 and f1-score 0.95 for children with CKD 4-5 on conservative treatment; precision 1.0, recall 0.86, and f1-score 0.92 for patients on chronic dialysis. The values of mean GINI importance measured for 40 random splits of the base for MPV, neutrophil count and SII, were 0.28, 0.32 and 0.39, respectively. Conclusion RFC built up with the input variables of neutrophil count, MPV, and SII, was the best predictor of progression into pediatric end stage kidney disease requiring chronic dialysis. SII turned out the most important parameter in the created model, having the largest share in the prediction of need for renal replacement therapy.

Clinical-radiomic models based on digital breast tomosynthesis images: a preliminary investigation of a predictive tool for cancer diagnosis.

This study aimed to develop a clinical-radiomic model based on radiomic features extracted from digital breast tomosynthesis (DBT) images and clinical factors that may help to discriminate between benign and malignant breast lesions. A total of 150 patients were included in this study. DBT images acquired in the setting of a screening protocol were used. Lesions were delineated by two expert radiologists. Malignity was always confirmed by histopathological data. The data were randomly divided into training and validation set with an 80:20 ratio. A total of 58 radiomic features were extracted from each lesion using the LIFEx Software. Three different key methods of feature selection were implemented in Python: (1) K best (KB), (2) sequential (S), and (3) Random Forrest (RF). A model was therefore produced for each subset of seven variables using a machine-learning algorithm, which exploits the RF classification based on the Gini index. All three clinical-radiomic models show significant differences (p < 0.05) between malignant and benign tumors. The area under the curve (AUC) values of the models obtained with three different feature selection methods were 0.72 [0.64,0.80], 0.72 [0.64,0.80] and 0.74 [0.66,0.82] for KB, SFS, and RF, respectively. The clinical-radiomic models developed by using radiomic features from DBT images showed a good discriminating power and hence may help radiologists in breast cancer tumor diagnoses already at the first screening.

MP17-15 INFRARED-(IR)-IMAGING CLASSIFIES PROSTATE CANCER LABEL-FREE: A PROSPECTIVE IR-PROSPECT-STUDY SHOWS POTENTIAL FOR CLINICAL APPLICATION AND RESEARCH

You have accessJournal of UrologyCME1 Apr 2023MP17-15 INFRARED-(IR)-IMAGING CLASSIFIES PROSTATE CANCER LABEL-FREE: A PROSPECTIVE IR-PROSPECT-STUDY SHOWS POTENTIAL FOR CLINICAL APPLICATION AND RESEARCH Henning Bahlburg, Claus Kuepper, Frederik Grosserueschkamp, Nina Goertzen, Carlo Sternemann, Sebastian Berg, Karl Tully, Axel Mosig, Rein-Jueri Palisaar, Andrea Tannapfel, Klaus Gerwert, Joachim Noldus, and Florian Roghmann Henning BahlburgHenning Bahlburg More articles by this author , Claus KuepperClaus Kuepper More articles by this author , Frederik GrosserueschkampFrederik Grosserueschkamp More articles by this author , Nina GoertzenNina Goertzen More articles by this author , Carlo SternemannCarlo Sternemann More articles by this author , Sebastian BergSebastian Berg More articles by this author , Karl TullyKarl Tully More articles by this author , Axel MosigAxel Mosig More articles by this author , Rein-Jueri PalisaarRein-Jueri Palisaar More articles by this author , Andrea TannapfelAndrea Tannapfel More articles by this author , Klaus GerwertKlaus Gerwert More articles by this author , Joachim NoldusJoachim Noldus More articles by this author , and Florian RoghmannFlorian Roghmann More articles by this author View All Author Informationhttps://doi.org/10.1097/JU.0000000000003237.15AboutPDF ToolsAdd to favoritesDownload CitationsTrack CitationsPermissionsReprints ShareFacebookLinked InTwitterEmail Abstract INTRODUCTION AND OBJECTIVE: The Goal of this study was to evaluate whether artificial intelligence (AI) is suited to classify prostate cancer (PCa) label-free. In the last decade, IR-microscopy was able to showcase its potential to classify different tissues and pathologies in several studies. Recorded IR-Spectra reflect the biochemical status of the examined cell. The application of quantum-cascade lasers as source of light has accelerated data acquisition and thus enabled integration into clinical research pathways METHODS: In this study, tissue from 258 patients was investigated. Initially, AI with Random Forrest (RF) algorithm was applied. Prostatic tissue showed to be morphologically more demanding than expected. As the study population increased, a robust deep-learning AI was implemented. RESULTS: Deep-learning AI reached an AUC of 0.96 with a sensitivity of ∼98% and a specificity of ∼83% when classifying PCa label-free. This marks an improvement compared to RF, which showed a sensitivity of 82% and a specificity of 69%. CONCLUSIONS: AI-based IR-Imaging detects PCa with a sensitivity of 98%. Further improvement may be achieved by additional training and increase in sample size. Subsequently, laser-assisted micro-dissection may allow precise and homogenous sampling for detailed molecular analysis of tumor tissues. Results may be used to train AI in the detection of molecular tumor characteristics such as oncological targets. Source of Funding: This project was funded by the Ministry of Culture and Research of the Federal State of Northrine Westphalia © 2023 by American Urological Association Education and Research, Inc.FiguresReferencesRelatedDetails Volume 209Issue Supplement 4April 2023Page: e218 Advertisement Copyright & Permissions© 2023 by American Urological Association Education and Research, Inc.MetricsAuthor Information Henning Bahlburg More articles by this author Claus Kuepper More articles by this author Frederik Grosserueschkamp More articles by this author Nina Goertzen More articles by this author Carlo Sternemann More articles by this author Sebastian Berg More articles by this author Karl Tully More articles by this author Axel Mosig More articles by this author Rein-Jueri Palisaar More articles by this author Andrea Tannapfel More articles by this author Klaus Gerwert More articles by this author Joachim Noldus More articles by this author Florian Roghmann More articles by this author Expand All Advertisement PDF downloadLoading ...

Deakin microgrid digital twin and analysis of AI models for power generation prediction

To achieve carbon neutral by 2025, Deakin University launched a AUD 23 million Renewable Energy Microgrid in 2020 with a 7-megawatt solar farm, the largest at an Australian University. A web-based digital twin (DT) is developed to provide operators with intelligence and insights through several AI-driven capabilities. Accurate and computationally efficient power generation prediction is one of the critical elements in this DT. To this end, we researched the literature and identified the commonly used Machine Learning-based prediction models and compared them computationally using power generation and weather sensor data obtained from the solar farm. From the computational experiments, we find that, overall, Artificial Neural Network (ANN) has achieved the highest R2-score (0.944) and the lowest RMSE (14.848). To obtain further insights, we compared the methods using our two novel metrics, the x-percentile Closeness scores and the x-percentile Absolute error scores. The new metrics provide us with a spectrum to measure the consistency and robustness of the prediction methods instead of just a single value. Further, power generation can fluctuate substantially, and a prediction model should be accurate regardless of the magnitude of the output, hence measuring the relative error has its merits. By our two new metrics, using the data from our Deakin Microgrid, Random Forrest (RF) outperformed the other methods tested, with the smallest absolute relative error across the whole spectrum (from 0.011 to 0.457). RF is also the fastest in model training time at 4.894 s and XGBoost came second at 5.115 s–a big contrast to ANN at 144.102 s. All prediction times are under 1 s. RF is therefore used as a power prediction algorithm in our Deakin Microgrid Digital Twin.

Abstract 11834: Ability of Machine Learning Algorithms to Predict Chronic Kidney Disease Using Health Examination Data From the Japanese General Population

Introduction: There is a large amount of specific health examination data in Japan. Efforts to utilize these data to predict the future onset of cardiovascular diseases are important from the perspective of improving prognosis and QOL as well as medical economy. We evaluated the prediction ability of machine learning algorithms for newly developed 5-year chronic kidney disease (CKD). Methods: Among those who underwent examinations from 2008 to 2018, individuals whose baseline (year X), 1 year later (year X+1), and 5 years later (year X+5) data were available were included. CKD was defined as estimated glomerular filtration rate (eGFR) &lt;60mL/min/1.73m2. Those with CKD in year X or X+1 were excluded. Three different machine learning algorithms, Random Forrest (RF), Logistic Regression (LR), and Multilayer Perceptron (MLP) to predict CKD on year X+5 from single-year data (year X) or consecutive 2 years data (year X and X+1) were assessed, and the predictive ability was compared. Anthropometric measurements, blood pressure and pulse rate, laboratory measurements, and lifestyle related indices were included as explanatory variables. Results: A total of 43,084 participants (age 53.7±10.1 years, male 49.9%) were included and 2,928 individuals (6.8%) had CKD in year X+5. The indices of predictive ability from single-year data were as follows: RF, AUC 0.88, accuracy 0.93, recall 0.99, precision 0.93; LR, AUC 0.89, accuracy 0.93, recall 0.99, precision 0.94; MLP, AUC 0.89, accuracy 0.93, recall 0.99, precision 0.93. Those from multi-year data were as follows: RF, AUC 0.93, accuracy 0.95, recall 0.99, precision 0.95; LR, AUC 0.90, accuracy 0.97, recall 0.99, precision 0.97; MLP, AUC 0.89, accuracy 0.97, recall 0.99, precision 0.97. The variables with the highest importance features were serum creatinine (0.41), age (0.13), medication for hypertension (0.04), and systolic blood pressure (0.03). Conclusions: The ability to predict new onset of CKD was better when using data from two consecutive years than when using data from a single year. The AUCs ranged from 0.89 to 0.93 and were generally better than previously reported risk models using statistical methods. Serum creatinine levels, along with age and blood pressure have been identified as major contributors.

Developing pedotransfer functions using Sentinel-2 satellite spectral indices and Machine learning for estimating the surface soil moisture

To estimate the surface soil moisture (SM) using a combination of new spectral indices and methods of Random Forrest (RF) and Multiple Linear Regression (MLR), 11 pedotransfer functions (PTF1-11) were developed by combining basic soil properties (clay, silt/sand, and bulk density) and spectral indices of Sentinel-2 satellite. In this study, 124 surface soil samples were randomly taken from three regions including Telo in Tehran province, Ivaneki in Semnan province, and Borujerd in Lorestan province, Iran. The results showed that the accuracy of the RF method was considerably higher compared to the MLR method. The SM was better estimated using water spectral indices (such as Normalized Difference Water Index (NDWI) and Surface Water Capacity Index (SWCI)), along with the basic properties of soil as inputs of PTF7. In the training and testing steps, the Root Mean Square Error (RMSE) decreased from 0.041 and 0.05 (cm3 cm−3) in PTF1 to 0.028 and 0.039 in PTF7, respectively. The average values of RMSE, Akaike Information Criterion (AIC), coefficient of determination (R2) and Relative Improvement (RI) of the RF method were 0.028 (cm3 cm−3), −559, 0.79 and 0.001, and 0.038 (cm3 cm−3), −279, 0.73 and −0.006 for the training and testing steps of all PTFs, respectively. While, these values for the MLR method were 0.032 (cm3 cm−3), −542, 0.69 and 0.0003, and 0.043 (cm3 cm−3), −269, 0.63 and 0.002 for the training and testing steps, respectively. Due to the low values of MBE, it was possible to disregard the overestimation of the results. The evaluation of results for predictor importance indicated that among the basic properties, clay percent has a significant effect on estimation of SM. These results show that spectral indices alone are not suitable estimators for SM estimation. It suggests using basic soil properties and spectral indices to estimate the SM.

Development of a Polymer Modified Quartz Tuning Fork (QTF) Sensor Array-Based Volatile Organic Compound (VOC) Classifier

Polymer modified Quartz Tuning Fork sensor array with three sensors has been developed to classify 1,4-dimethoxy-2, 3-butanediol (BD), cyclohexanone (CH), ethanol, methanol and acetone with various concentrations. Linear Discriminant Analysis (LDA), Decision Tree (DT), and Random Forrest (RF) have been used and evaluated as the classifiers predicting the VOCs and their concentration. The classifiers were trained with three features: response time, recovery time, and shift in the resonant frequency of the sensors. The results have been compared and analyzed for the performance of each classifier for the present system. RF gives the best performance as a classifier with an accuracy of classification more than 95%. We present results of VOC classifier performance for the QTF sensor array.

An evolutionary approach to comparative analysis of detecting Bangla abusive text

The use of Bangla abusive texts has been accelerated with the progressive use of social media. Through this platform, one can spread the hatred or negativity in a viral form. Plenty of research has been done on detecting abusive text in the English language. Bangla abusive text detection has not been done to a great extent. In this experimental study, we have applied three distinct approaches to a comprehensive dataset to obtain a better outcome. In the first study, a large dataset collected from Facebook and YouTube has been utilized to detect abusive texts. After extensive pre-processing and feature extraction, a set of consciously selected supervised machine learning classifiers i.e. multinomial Naïve Bayes (MNB), multi layer perceptron (MLP), support vector machine (SVM), decision tree, random forrest, stochastic gradient descent (SGD), ridge, perceptron and k-nearest neighbors (k-NN) has been applied to determine the best result. The second experiment is conducted by constructing a balanced dataset by random under sampling the majority class and finally, a Bengali stemmer is employed on the dataset and then the final experiment is conducted. In all three experiments, SVM with the full dataset obtained the highest accuracy of 88%.

A machine learning framework for predicting long-term graft survival after kidney transplantation

Kidney transplantation (KT) is an optimal treatment for end-stage renal disease (ESRD). Currently, short-term KT outcomes are indeed excellent, but long-term successful outcomes are still difficult to achieve, and improving them is crucial for kidney recipients. An early and accurate prediction of long-term graft survival helps healthcare practitioners to create a more personalized treatment plans for patients and facilitates the performance of clinical trials. In this study, we propose a machine learning framework to early predict graft survival after five years of KT and determine the most influential parameters that affect the survival. Our dataset was collected from Charles Nicolle Hospital in Tunis in Tunisia and it included pre, peri, post KT aspects. We utilized four machine learning algorithms to select the most important features: the least absolute shrinkage and selection operator logistic regression (Lasso-LR), Random Forrest (RF), Decision Tree (DT), and Chi-square (Chi-sq). We utilized three Scikit-learn functions to implement those algorithms: SelectFromModel (SFM), Recursive Feature Elimination (RFE), and SelectKBest (SKB). Five algorithms were utilized to builds prediction models based on the data groups resulted from the feature selection step: logistic regression (LR), k-nearest neighbors (KNN), extreme gradient boosting (XGB), and artificial neural network (ANN). We evaluated the models using five performance measures: accuracy, sensitivity, specificity, F1 measure, and area under the curve (AUC). XGBoost resulted the best model with the highest AUC (89.7%). It was based ten features selected by RF algorithm and SFM function. The accuracy, sensitivity, specificity, and F1 of the best model were 91.5%, 91.9%, 87.5%, and 89.6%, respectively. This study proposes a novel approach for investigating long-term allograft survival while considering the complex relationship between all KT aspects and long-term outcomes. Our framework can be used as a decision support system for Nephrologists to early detect graft status, which helps in developing safer recommendations for kidney patients and consequently obtaining positive KT outcomes and mitigating the risks of graft failure.

Data-Driven Prediction of Freezing of Gait Events From Stepping Data.

Freezing of gait (FoG) is typically a symptom of advanced Parkinson's disease (PD) that negatively influences the quality of life and is often resistant to pharmacological interventions. Novel treatment options that make use of auditory or sensory cues might be optimized by prediction of freezing events. These predictions might help to trigger external sensory cues—shown to improve walking performance—when behavior is changed in a manner indicative of an impending freeze (i.e., when the user needs it the most), rather than delivering cue information continuously. A data-driven approach is proposed for predicting freezing events using Random Forrest (RF), Neural Network (NN), and Naive Bayes (NB) classifiers. Vertical forces, sampled at 100 Hz from a force platform were collected from 9 PD subjects as they stepped in place until they at least had one freezing episode or for 90 s. The F1 scores of RF/NN/NB algorithms were computed for different IL (input to the machine learning algorithm), and GL (how early the freezing event is predicted). A significant negative correlation between the F1 scores and GL, highlighting the difficulty of early detection is found. The IL that maximized the F1 score is approximately equal to 1.13 s. This indicates that the physiological (and therefore neurological) changes leading to freezing take effect at-least one step before the freezing incident. Our algorithm has the potential to support the development of devices to detect and then potentially prevent freezing events in people with Parkinson's which might occur if left uncorrected.

RICE CROP MAPPING USING SENTINEL-1 TIME SERIES IMAGES (CASE STUDY: MAZANDARAN, IRAN)

Abstract. Policymaking and planning agricultural improvement require accurate and timely information and statistics. In Iran, collecting and acquiring agricultural statistics is often done in the traditional methods. Related studies have proved that these methods mostly contain some mistakes. Multi-temporal acquisition strategies of remotely sensed data provide an opportunity to improve rice monitoring and mapping. Studying and monitoring rice paddies in vast areas is limited by the presence of cloud cover, the spatial and temporal resolution of optical sensors, and the lack of open access or systematic Radar data. Sentinel-1 satellite data, which are free to access and has a high quality of spatial and temporal resolution, can provide a great opportunity for monitoring crop products, especially rice. In this study, Sigma Nought, Gamma Nought and Beta Nought time series of Sentinel-1 data in VV, VH and VV+VH polarizations were employed for extracting areas under rice cultivation in the region of Mazandaran province, Iran. These satellite data are taken regularly every 12 days, according to the season of the region, from March 21st to September 22nd of 2018. In this study, in order to specify the rice paddies area, several fieldworks were randomly carried out for two weeks, and field data were collected as well. Field data including rice paddies areas and non-rice areas were collected as ‘Test and Train data set’ and then the Random Forrest (RF) algorithm was carried out to determine the rice paddies area. The classification result was validated using test samples. The accuracy of all classifications results are over 80% and the best result is related to Sigma Nought and gamma Nought of VH polarization, with an accuracy of 91.37%. The results showed a high capability to evaluate and monitor rice production at moderate levels in a vast area which is regularly exposed to the cloud cover.

A Novel Co-Training-Based Approach for the Classification of Mental Illnesses Using Social Media Posts

Context: Recently, research community of certain domain showing their eagerness towards the use of social media networks to gain constructive knowledge in decision making and automation, such as aid to perform software development activities, crypto-currencies usage, network community detection and recommendation and so on. Recently, besides other domains of eHealth, the use of social media and big data analytics has become hot topic to predict the patient of mental illness involved in either depression, schizophrenia, eating disorders, anxiety or addictive behaviors. Problem: Traditional methods either need enough historic data or to keep the regular monitoring on patient activities for identification of a patient associated with a mental illness disease. Method: In order to address this issue, we propose a methodology to classify the patients associated with chronic mental illness diseases (i.e. Anxiety, Depression, Bipolar, and ADHD (Attention Deficit Hyperactivity Disorder) based on the data extracted from the Reddit, a well-known network community platform. The proposed method is employed through Co-training (type of semi-supervised learning approach) technique by incorporating the discriminative power of widely used classifiers namely Random Forrest (RF), Support Vector Machine (SVM), and Naive Bayes (NB). We used Reddit API to download posts and top five associated comments for construction of a feature space. Results: The experimental results indicate the effectiveness of Co-training based classification rather than the state of the art classifiers by a margin of 3% on average in par with every state of art technique. In future, the proposed method could be employed to investigate any classification problem of any domain by extracting date from the social media.

Modify Random Forest Algorithm Using Hybrid Feature Selection Method

The Importance of Random Forrest(RF) is one of the most powerful ‎methods ‎of ‎machine learning in ‎Decision Tree.‎ The Proposed hybrid feature selection for Random Forest depend on ‎two ‎measure ‎‎Information Gain and Gini Index in varying percentages ‎based on ‎weight.‎ In this paper, we tend to ‎propose a modify Random Forrest‏ ‏‎algorithm named ‎Random Forest algorithm using hybrid ‎feature ‎‎selection ‎that uses hybrid feature ‎selection instead of ‎using ‎one feature selection. The ‎main plan is to ‎computation the ‎‎ Information ‎Gain for all random selection ‎feature then search for ‎the best split ‎‎point in ‎the node that gives the best ‎value for a hybrid ‎equation with ‎Gini Index. ‎The experimental results on the ‎dataset ‎showed that the proposed ‎modification is ‎better than the classic Random ‎Forest compared to ‎the standard static Random ‎Forest the hybrid feature ‎‎selection Random Forrest shows significant ‎improvement ‎in accuracy measure.‎

Identification and Validation of Sensory-Active Compounds from Data-Driven Research: A Flavoromics Approach.

In this study, highly predictive LC-MS features (retention time_ m/ z) derived from untargeted chemical fingerprinting-multivariate analysis (MVA) previously used to model flavor changes in citrus fruits related to aging (freshness) were further isolated and analyzed for sensory impact, followed by structural elucidation. The top 10 statistical features from two MVA approaches, partial least-squares data analysis (PLS-DA) and Random Forrest (RF), were purified to approximately 70% via multidimensional liquid chromatography-mass-directed fractionation to screen for sensory activity. When added to a 'fresh' orange flavor model system, 50-60% of the isolates were reported to cause a sensory change. From the subset of the actives identified, two compounds were selected, on the basis of statistical relevance, that were further purified to >97% for identification (MS, NMR) and for sensory descriptive analysis (DA). The compounds were identified as nomilin glucoside and a novel ionone glucoside. DA evaluation in the recombination orange model indicated both compounds statistically suppressed the perceived intensity of the "orange character" attribute, whereas the novel ionone glycoside also decreased the intensity of the floral character while increasing the green bean attribute intensity.

Mode Choice Analysis Using Random Forrest Decision Trees

Mode choice analysis forms an integral part of transportation planning process as it gives a complete insight to the mode choice preferences of the commuters and is also used as an instrument for evaluation of introduction of new transport systems. Mode choice analysis involves the procedure to study the factors in decision making process of the commuter while choosing the mode that renders highest utility to them. This study aims at modelling the mode choice behaviour of commuters in Delhi by considering Random Forrest (RF) Decision Tree (DT) method. The random forest model is one of the most efficient DT methods for solving classification problems. For the purpose of model development, about 5000 stratified household samples were collected in Delhi through household interview survey. A comparative evaluation has been carried out between traditional Multinomial logit (MNL) model and Decision tree model to demonstrate the suitableness of RF models in mode choice modelling. From the result, it was observed that model developed by Random Forrest based DT model is the superior one with higher prediction accuracy (98.96%) than the Logit model prediction accuracy (77.31%).

Detector-C: A blood-based IVD with high sensitivity and specificity for early detection of colorectal cancer.

3580 Background: Colonoscopy is the gold standard for screening of colorectal cancer (CRC). However, only 20%-30% of the screening population aged 55-75 years use colonoscopy due to its inconvenience and invasive nature. There is a need for an accurate, noninvasive blood test. Methods: We randomized 572 PAX-RNA blood samples from two prospective, multicentre clinical studies. CRC cases originated from the MSKK study, controls from the CRC.SCR.2 study which enrolled healthy individuals from a screening population undergoing colonoscopy. Blood was drawn prior to surgery or prior to colonoscopy. RNA was isolated, QC checked, hybridized onto the Affymetrix U133 2.0 Plus arrays, and QC checked. Clinical data were monitored. Results: After all QCs data of 462 pts remained. Data from 119 pts (55 cases, 64 controls) were utilized for discovery of gene signatures using a double nested bootstrap with random forrest and SVM as feature selection methods. The best signature, Detector-C, discriminating between cases and controls contained 202 probe sets and resulted in a prospective sensitivity (S+) of 85.3% and a prospective specificity (S-) of 93.3%. The hypothesis for the prospective performance evaluation (validation) of Detector-C was to show a S+ of > 70% and a S− of > 80% with a one-sided alpha of 2.5% and a power of 97.5%. In order to test this hypothesis at least 122 cases and at least 109 controls would be needed. We used the remaining 210 cases and 133 controls for the prospective validation. S+ was 90% (95% CI: 0.851-0.937) and S− was 88% (95% CI: 0.812-0.930). S+ by UICC stage was: stage I: 0.89 (95% CI: 0.774-0.958), stage II: 0.90 (95% CI: 0.788-0.961), stage III: 90% (95% CI: 0.805-0.959) and stage IV: 93% (95% CI: 0.765-0.991). Detector-C also identified high-grade intraepithelial neoplasia with S+ of 66%. Multivariate analysis showed no significant effect of stage, age, gender, tumor localization, or RNA quality on correct prediction. Conclusions: Detector-C measures the host response (host defense, immune answer, wound healing, inflammation) of white blood cells to tumor lesions and is highly suited for screening of CRC due its easy of use, noninvasiveness and its high sensitivity for early-stage CRC/neoplastic lesions. Author Disclosure Employment or Leadership Position Consultant or Advisory Role Stock Ownership Honoraria Research Funding Expert Testimony Other Remuneration Signature Diagnostics AG Signature Diagnostics AG Signature Diagnostics AG Signature Diagnostics AG