Prediction Of Flight Delays Using Feature Engineering, Catboost, And Bayesian Optimization To Improve Model Performance

Flight delay is a major issue in the aviation industry. In commercial aviation, if a flight reaches its destination 15 minutes later than the scheduled arrival, it is said to be delayed. Flight delays cause a great deal of bother to travelers. It could make them late to their booked occasions or miss a corresponding flight, accordingly prompting outrage and dissatisfaction. Likewise, travelers may not generally be entitled for a refund when a postponement happens. Carriers report that couple of the numerous reasons prompting most flight delays are carrier glitches, climate conditions, support issues with the airplane and congestion in air traffic. Rapid development in airline industry has led to an increased number of aircrafts in the skies, this has brought about air-gridlock causing flight delays. Flight delays are not only extremely undesirable financially but also have adverse environmental effects. Air-traffic management is becoming increasingly challenging. The aim of our research work is to predict the delay of flights due to various factors using machine learning and deep learning so as to minimize losses and increase customer satisfaction. This Machine Learning model could be integrated with Airlines systems for the use of staff and customers and it could also rank Airlines and flights based on delays.

Objective To evaluate the six scoring systems and four laboratory tests, including pancreatitis outcome prediction (POP), Ranson score, bedside index for severity in acute pancreatitis (BISAP), acute physiology and chronic health evaluation Ⅱ (APACHEⅡ), systemic inflammatory response syndrome (SIRS), and Glasgow score as well as four laboratory tests including C-reactive protein (CRP), hematocrit (HCT), blood urea nitrogen (BUN) and serum creatinine (Scr) in the prognostic assessment of severe acute pancreatitis (SAP). Methods From January 2016 to December 2017, at Sir Run Run Shaw Hospital, 151 SAP patients who met the enrollment criteria were retrospectively analyzed. According to the time from onset to treatment, the patients were divided into less than three days group (n=102) and over three days group (n=49). The evaluation of six scoring systems and four laboratory tests, including CRP, HCT, BUN and Scr at 0, 24 and 48 h after hospitalization in the prognostic assessment of SAP patients was measured by receiver-operating characteristic (ROC) curve. Results The Ranson score had the highest area under curve (AUC) value (0.916) in the evaluation of the prognosis of SAP patients less than three days group followed by BISAP, APACHEⅡ, Glasgow and POP score, and their AUC values were 0.832, 0.823, 0.793, and 0.787, respectively, all of them were statistically significant in the prognostic assessment of SAP patients in less than three days group (all P<0.05). There were statistically significant of BISAP and APACHEⅡ scores in the prognostic evaluation of SAP patients in over three days group (both P<0.05), and the AUC values were 0.751 and 0.735, respectively, which were less than those of SAP patients in less than three days group. There were statistical significance of BUN and Scr at 24 and 48 h after hospitalization in the prognostic assessment of SAP patients in less than three days group (all P<0.05), and the AUC values were 0.856, 0.853 and 0.793, 0.874, respectively. There were statistical significance of BUN at 0, 24, 48 h and Scr at 48 h after hospitalization in the prognostic assessment of SAP patients in over three days group (all P<0.05), and the AUC value was 0.709, 0.754, 0.742 and 0.716, respectively. Conclusions Ranson, POP and Glascow score systems are only suitable for patients with SAP less than three days. APACHEⅡ, BISAP score systems, BUN and Scr can be used to evaluate patients with SAP over three days, but are more suitable for patients with SAP less than three days group. Key words: Severe acute pancreatitis; pancreatitis of outcome prediction; Ranson score; BISAP; APACHE Ⅱ; systemic inflammatory response syndrome; Glasgow; ROC curve

1113: The Relationship Between Tumor Perfusion Parameters on Contrast-Enhanced Ultrasonography and Differentiation of Hepatocellular Carcinomas

Delays in flights and other airline operations have significant consequences in quality of service, operational costs, and customer satisfaction. Therefore, it is important to predict the occurrence of delays and take necessary actions accordingly. In this study, we addressed the flight delay prediction problem from a supervised machine learning perspective. Using a real-world airline operations dataset provided by a leading airline company, we identified optimum dataset features for optimum prediction accuracy. In addition, we trained and tested 11 machine learning models on the datasets that we created from the original dataset via feature selection and transformation. CART and KNN showed consistently good performance in almost all cases achieving 0.816 and 0.807 F-Scores respectively. Similarly, GBM, XGB, and LGBM showed very good performance in most of the cases, achieving F-Scores around 0.810.

Nowadays, the civil aviation industry has a high precision demand of flight delay prediction. To make full use of the characteristics of flight data and meteorological data, two flight delay prediction models using deep convolution neural network based on fusion of meteorological data are proposed in this paper. One is DCNN (Dual-channel Convolutional Neural Network), which refers to the ResNet network structure. The other is SE-DenseNet (Squeeze and Excitation-Densely Connected Convolutional Network), combining the advantages of DenseNet and SENet. Firstly, flight data and meteorological data are fused in the model. Then, both DCNN and SE-DenseNet models are used to extract feature automatically based on the fused flight data set. Finally, the softmax classifier is adopted to predict the flight delay level. For proposed DCNN model, both straight channel and convolution channel are designed to guarantee the lossless transmission of the feature matrix and enhance the patency of the deep network. For proposed SE-DenseNet model, a SE module is added after the convolution layer of each DenseNet block, which can not only enhance the transmission of deep information but also achieve feature recalibration in the feature extraction process. The research results indicate that after considering characteristics of meteorological information, the accuracy of the model can be improved 1% compared with only considering the flight information. The two deep convolutional neural networks proposed in this paper, DCNN and SE-DenseNet, can both effectively improve the prediction accuracies, reaching to 92.1% and 93.19%, respectively.

The purpose of this paper is to combine the cloud model method with the Artificial neural network to predict and evaluate flight delay date. In this paper, the flight delay data of the United States in recent 15 years were collected, involving nine airlines and seven delay causes. Then, the flight delay prediction model is obtained by combining neural network and cloud model theory. 36 sets of data are extracted to verify the accuracy of the prediction model. Finally, the delay rate is predicted and the prediction data is evaluated by cloud model. The results show that the amount of data to be processed is reduced to 20 % after combining the cloud model with the neural network. In the model validation, the average expected value deviation of flight delay prediction model is 23.1 %. In the analysis and evaluation of the prediction results, the delay cause 6 is taken as an example, with the highest delay rate of 7.03 % in December. The delay risk assessment results for all months were general. The method provided in this paper can predict flight delays and provide theoretical support for the delay prevention of airlines.

By predicting flight delays, economic losses caused by delay situations can be effectively prevented and reduced. A classification algorithm for flight delay prediction based on unbalanced data is proposed to solve the problem of unbalanced number of delayed and non-delayed samples in the process of flight delay prediction. Firstly, a multi-classification model based on the CycleMLP algorithm is proposed for flight delay prediction. Secondly, the CycleMLP algorithm is improved by using the Focal Loss modified cross-entropy loss function to build the final classification model for flight delay prediction. And the experimental results show that the improved model has better prediction results for delayed flights, with a 5.54% increase in average macro precision and 1.86% increase in accuracy. Finally, the flight delay prediction results are published in the mobile APP application system to provide corresponding advice to passengers and airport departments.

Accurate flight delay prediction contains great reference value for airline business and passenger travel. Recent studies have been concentrated on applying machine learning methods to predict the probability of flight delay. Most of the previous prediction methods are built for a single air route or airport. This paper explores a broader spectrum of factors that may potentially affect the flight delay and proposes a gradient boosting decision tree (GBDT) based models for generalized flight delay prediction. To build a dataset for the proposed model, automatic dependent surveillance-broadcast (ADS-B) messages are received, pre-processed, and integrated with other information such as weather condition of airport, flight schedule, and airport information. Since the delay prediction results can be given with higher resolution, the designed prediction tasks contain four different classification tasks. Experimental results show that the proposed GBDT-based model can obtain higher prediction accuracy (87.72% for the binary classification) when handling our limited dataset.

e20590 Background: There are no blood-based biomarkers for survival prediction in inoperable stage III NSCLC patients. We propose a method for calculation of novel area under curve (AUC) biomarkers of the dynamic change of multiple leukocyte subpopulations before, during, and after thoracic irradiation (TRT), chemoradiotherapy (CRT), and chemo-radio-immunotherapy (CRT-ICI) in this patient cohort. The extracted biomarkers identify patients with early progression after TRT. Methods: 20 patients (17 male, 3 female), at median age of 65.5 years (range 34 to 79) were enrolled in the study. The median follow-up time was 60 weeks. Eleven patients suffered from adenocarcinoma, 8 squamous cell carcinoma, and 1 undifferentiated NSCLC. They received TRT (2/20, 10%), CRT (11/20, 55%), or CRT-ICI (7/20, 35%). One patient (1/20, 5%) withdrew consent and was excluded from analysis. Primary endpoints were progression free survival (PFS) at 6 and 12 months. Patient blood was analyzed via flow cytometry for 7 circulating leukocyte populations at baseline, twice during radiotherapy, at the end of radiotherapy (RTend), and 10, 20, 35, 48, and 60 weeks after enrollment. We analyzed CD3+ total T cells, CD4+ T cells, CD8+ T cells, CD19+/CD20+ B cells, CD3-CD56+ NK cells, CD56 bright NK cells, and CD56 dim NK cells. Here, we report on CD4+ and CD8+ T cells. The AUC from RTend to zenith of absolute cell counts provided aggregate measures for the time-course data. We performed hierarchical clustering and cluster characterization. Relevant features were selected by stepwise drop-out. Results: Clustering of the AUC between RTend and zenith for CD4+ T cells and CD8+ T cells delineated two prognostic groups. The favorable group was characterized by higher AUC values for CD4+ and CD8+ T cells compared to the unfavorable group. All patients (9/9, 100%) in the favorable group versus 36.4% (4/11) patients in the unfavorable group were progression-free at 6 months (Fisher´s exact test, two-tailed: p-value = 0.00472). This effect was observed as a trend with PFS at 12 months. Here 66.6% (6/9) of patients had PFS at 12 months in the favorable group versus 27.3% (3/11) in the unfavorable group (Fisher´s exact test, two-tailed: p-value = 0.175). There is a directly proportional relationship of the reported AUC values to PFS at 6 months and 12 months. Conclusions: Patients who responded with T cell expansion after immunogenic cell death by TRT, CRT, or CRT-ICI had significantly longer PFS compared to those without increase. Longitudinal monitoring of CD4+ and CD8+ T cells and the AUC from RTend to zenith is a promising biomarker for detecting early progression in the present study which is the subject of validation in an ongoing prospective study (PRECISION, NCT05027165).

This study aimed to determine thresholds for percentage of body fat (BF%) corresponding to the cut-off values for overweight/obesity as recommended by the International Obesity Task Force (IOTF), using two bioelectrical impedance analyzers (BIA), and described the likelihood of increased cardiometabolic risk in our cohort defined by the IOTF and BF% status. Participants included 1165 children and adolescents (54.9% girls) from Bogotá (Colombia). Body mass index (BMI) was calculated from height and weight. BF% of each youth was assessed first using the Tanita BC-418® followed by a Tanita BF-689®. The sensitivity and specificity of both devices and their ability to correctly classify children as overweight/obesity (≥2 standard deviation), as defined by IOTF, was investigated using receiver operating characteristic (ROC) by sex and age groups (9–11, 12–14, and 13–17 years old); Area under curve (AUC) values were also reported. For girls, the optimal BF% threshold for classifying into overweight/obesity was found to be between 25.2 and 28.5 (AUC = 0.91–0.97) and 23.9 to 26.6 (AUC = 0.90–0.99) for Tanita BC-418® and Tanita BF-689®, respectively. For boys, the optimal threshold was between 16.5 and 21.1 (AUC = 0.93–0.96) and 15.8 to 20.6 (AUC = 0.92–0.94) by Tanita BC-418® and Tanita BF-689®, respectively. All AUC values for ROC curves were statistically significant and there were no differences between AUC values measured by both BIA devices. The BF% values associated with the IOTF-recommended BMI cut-off for overweight/obesity may require age- and sex-specific threshold values in Colombian children and adolescents aged 9–17 years and could be used as a surrogate method to identify individuals at risk of excess adiposity.

With the development of the civil aviation transportation industry in recent years, the volume of civil aviation transportation has increased rapidly. Increased carrier costs and reduced airport operating efficiency caused by flight delays have become issues that need to be addressed. How to improve the accuracy of predicting flight arrival delay time is of great significance for improving airport transportation efficiency, rationally scheduling flights and improving passenger comfort. In this paper, the Cat-boost model is utilized on the U.S Domestic airline on-time performance data from U.S. Transportation Administration, combined with the characteristics of the model to determine the influencing factors, and to predict the arrival delays of flights within the United States. The accuracy;precision and some other criterion of the model are given to evaluate the performance on the data. A better effect is obtained: the accuracy reach 80.44% in this case. Finally, the specific delay time is predicted, we found that the support vector machine has the best prediction result for the flight delay time, the average prediction error is 9.733 min, which has a certain reference value for flight operation and airport scheduling.

Features engineered hardness and yield strength prediction of septenary refractory high-entropy alloys

Bayesian optimization (BO) is a versatile and robust global optimization method under uncertainty. However, most of the BO algorithms were developed for problems with only continuous variables. For practical engineering optimization, discrete variables are also prevalent. BO methods based on Gaussian process (GP) surrogates also suffers from the curse-of-dimensionality problem. To address these challenges, in this paper, a Bayesian discrete optimization algorithm is introduced to solve permutation-based combinatorial problems. A new kernel function is developed based on position distances for permutation. To improve the efficiency and scalability of the algorithm, a sparse GP model based on inducing points is further developed, where the simulated annealing algorithm is applied to select inducing points. The new algorithm is demonstrated and tested with the production scheduling problem for additive manufacturing. Experimental results show that the proposed algorithm can find a better solution with limited evaluations than state-of-the-art algorithms.

BackgroundRecently, cancer organoid-based drug sensitivity tests have been studied to predict patient responses to anticancer drugs. The area under curve (AUC) or IC50 value of the dose-response curve (DRC) is used to differentiate between sensitive and resistant patient‘s groups. This study proposes a multi-parameter analysis method (cancer organoid-based diagnosis reactivity prediction, CODRP) that considers the cancer stage and cancer cell growth rate, which represent the severity of cancer patients, in the sensitivity test.MethodsOn the CODRP platform, patient-derived organoids (PDOs) that recapitulate patients with lung cancer were implemented by applying a mechanical dissociation method capable of high yields and proliferation rates. A disposable nozzle-type cell spotter with efficient high-throughput screening (HTS) has also been developed to dispense a very small number of cells due to limited patient cells. A drug sensitivity test was performed using PDO from the patient tissue and the primary cancer characteristics of PDOs were confirmed by pathological comparision with tissue slides.ResultsThe conventional index of drug sensitivity is the AUC of the DRC. In this study, the CODRP index for drug sensitivity test was proposed through multi-parameter analyses considering cancer cell proliferation rate, the cancer diagnosis stage, and AUC values. We tested PDOs from eight patients with lung cancer to verify the CODRP index. According to the anaplastic lymphoma kinase (ALK) rearrangement status, the conventional AUC index for the three ALK-targeted drugs (crizotinib, alectinib, and brigatinib) did not classify into sensitive and resistant groups. The proposed CODRP index-based drug sensitivity test classified ALK-targeted drug responses according to ALK rearrangement status and was verified to be consistent with the clinical drug treatment response.ConclusionsTherefore, the PDO-based HTS and CODRP index drug sensitivity tests described in this paper may be useful for predicting and analyzing promising anticancer drug efficacy for patients with lung cancer and can be applied to a precision medicine platform.

To prospectively compare the diagnostic performances of three methods of region of interest (ROI) placement for the measurements of intravoxel incoherent motion (IVIM) diffusion-weighted MR imaging in differentiating the histologic grade of hepatocellular carcinoma (HCC). Eighty-seven patients with 91 newly diagnosed HCCs were studied using IVIM imaging. Two attending radiologists separately identified the selection of tumour tissue for ROI positioning. Three different ROI positioning methods, namely the whole tumour volume (WTV) method, three-ROI method and one-section method, were used for the measurement. Kruskal-Wallis rank test or one-way ANOVA was used to compare the difference in IVIM parameters and ADC across the three different ROI positioning methods. Spearman correlation analysis was used to determine the correlation between each parameter and Edmondson-Steiner (E-S) grade. Receiver operating characteristics (ROC) curve analyses were performed to evaluate the diagnostic performance. For the ADC and ADCslow, the mean value measured by using the WTV method was significant higher than the one-section and three-ROI methods (all p < 0.01). For the ADCslow, the highest area under curve (AUC) with a value of 0.969 was obtained by using the WTV method, followed by the one-section method (AUC = 0.938) and three-ROI method (AUC = 0.873). Additionally, for the ADC, AUC values were 0.861 for WTV method, 0.840 for one-section method and 0.806 for three-ROI method. Different ROI positioning methods used significantly affect the IVIM parameters and ADC measurements. Measurements of ADCslow value derived from WTV method entailed the highest diagnostic performance in grading HCC. • Diffusion MRI is useful for non-invasively differentiating the histologic grade of hepatocellular carcinoma. • Different ROI positioning methods used significantly affect the IVIM parameters and ADC measurements. • IVIM model is advantageous over mono-exponential model for assessing the histologic grade of hepatocellular carcinoma.