Forest Algorithm Research Articles

Application of the Sino US Financial Technology in Banks

Objectives: With the rise and development of Internet finance, the application of Sino US financial technology in the banking field is becoming more and more widely. Methods: In this study, for the data collection of bank customer deposits, data mining and decision tree analysis algorithms were used to classify bank customers. Results: The classification accuracy of the traditional algorithm was low, so the optimization algorithm Adaboost and the random forest improvement algorithm were proposed in this paper. The simulation effects of its application in data combination show that the classification effect of the optimization algorithm is obviously better than the traditional classification algorithm. Conclusion: The results of this study can help banks gain customers and reduce expenditures.

Mbl-2 gene polymorphisms in pediatric Burkitt lymphoma: an approach based on machine learning techniques

Introduction: Burkitt lymphoma belongs to the group of non-Hodgkin lymphomas. Although curable in 80% of less advanced stages, it presents in advanced stages in about 75% of cases in Brazil’s Northeast region, requiring urgent and intensive care in the early stages of treatment. Objectives: therefore, this study aimed to verify the participation of MBL-2 gene polymorphisms in the development of Burkitt lymphoma. Methods: In this article, computational approaches based on the Machine Learning technique were used, where we implemented the Random Forest and KMeans algorithms to classify patterns of individuals diagnosed with the disease and, therefore, differentiate them from healthy individuals. A group of 56 patients aged 0 to 18 years, with Burkitt lymphoma, from a reference hospital in the treatment of childhood cancer, was evaluated, together with a control group consisting of 150 samples, all of which were tested for exon 1 polymorphisms and the MBL2 gene -221 and -550 regions. Results: At first, an unsupervised classification was performed, which identified as two the number of groups that best represent the data present in our database, reaching 72.81% accuracy in the separation of patients and controls. Then, the supervised classification was performed, where the classifier obtained a 70.97% success rate, being possible to reach 75% accuracy in the best GridSearch configuration when performing a cross validation. Conclusion: It was not yet possible to conclude about the participation of the evaluated polymorphisms in the development of the BL, however the computational techniques used proved to be very promising for carrying out studies of this nature.

Spatial modelling of accidents risk caused by driver drowsiness with data mining algorithms

Driver drowsiness causes many road accidents, and preparing a risk map of these accidents with spatial criteria and data mining algorithms highlights accident points well. In this study, accidents risk caused by driver drowsiness in Qazvin province, Iran, was modelled using decision tree (DT), random forest (RF) and support-vector regression (SVR) algorithms in GIS environment. Seven spatial criteria including road segment length, road width, slope angle, speed limit, land use/cover, distance to service area and distance to speed camera were selected as effective criteria in modelling. The effect of criteria in modelling was applied using a fuzzy method, and three risk maps were prepared. Evaluation with ROC-AUC showed that the AUC for RF, SVR and DT models were 0.904, 0.863 and 0.805, respectively, and the RF model overall had the best performance. Examining the importance of criteria showed that the speed limit was the most important criterion for modelling.

Development of an Adaptive Model for the Rate of Steel Corrosion in a Recirculating Water System

The stable quality of circulating water ensures the long-term stable operation of various processes in petrochemical production and achieves energy savings and emission reduction while reducing environmental pollution and yielding economic benefits to petrochemical enterprises. However, traditional circulating water quality evaluation and modeling for corrosion rate prediction suffer from adaptability and accuracy problems. To address these problems, the water quality analysis data of the circulating water in the field were subjected to data preprocessing and water quality index calculation to perform feature engineering, followed by modeling using a machine learning method that integrates the adaptive immune genetic algorithm and random forest (RF) algorithm and can intelligently select the water quality parameters to be used as the input variables for the RF modeling. Finally, the method was validated using an industrial example, and the results indicate that the method is capable of removing interference variables and is suitable for carbon steel corrosion rate prediction based on water quality models. The proposed method provides a basis for water quality management and real-time decision-making by circulating water field personnel.

Real-Time AI-Based Informational Decision-Making Support System Utilizing Dynamic Text Sources

Unstructured data from the internet constitute large sources of information, which need to be formatted in a user-friendly way. This research develops a model that classifies unstructured data from data mining into labeled data, and builds an informational and decision-making support system (DMSS). We often have assortments of information collected by mining data from various sources, where the key challenge is to extract valuable information. We observe substantial classification accuracy enhancement for our datasets with both machine learning and deep learning algorithms. The highest classification accuracy (99% in training, 96% in testing) was achieved from a Covid corpus which is processed by using a long short-term memory (LSTM). Furthermore, we conducted tests on large datasets relevant to the Disaster corpus, with an LSTM classification accuracy of 98%. In addition, random forest (RF), a machine learning algorithm, provides a reasonable 84% accuracy. This research’s main objective is to increase the application’s robustness by integrating intelligence into the developed DMSS, which provides insight into the user’s intent, despite dealing with a noisy dataset. Our designed model selects the random forest and stochastic gradient descent (SGD) algorithms’ F1 score, where the RF method outperforms by improving accuracy by 2% (to 83% from 81%) compared with a conventional method.

SCAFFY

Denial of service (DoS) attack is one of the common threats to the availability of critical infrastructure and services. As more and more services are online enabled, the attack on the availability of these services may have a catastrophic impact on our day-to-day lives. Unlike the traditional volumetric DoS, the slow DoS attacks use legitimate connections with lesser bandwidth. Hence, it is difficult to detect slow DoS by monitoring bandwidth usage and traffic volume. In this paper, a novel machine learning model called ‘SCAFFY' to classify slow DoS on HTTP traffic using flow level parameters is explained. SCAFFY uses a multistage approach for the feature section and classification. Comparison of the classification performance of decision tree, random forest, XGBoost, and KNN algorithms are carried out using the flow parameters derived from the CICIDS2017 and SUEE datasets. A comparison of the result obtained from SCAFFY with two recent works available in the literature shows that the SCAFFY model outperforms the state-of-the-art approaches in classification accuracy.

Digital mapping of soil texture in ecoforest polygons in Quebec, Canada.

Texture strongly influences the soil’s fundamental functions in forest ecosystems. In response to the growing demand for information on soil properties for environmental modeling, more and more studies have been conducted over the past decade to assess the spatial variability of soil properties on a regional to global scale. These investigations rely on the acquisition and compilation of numerous soil field records and on the development of statistical methods and technology. Here, we used random forest machine learning algorithms to model and map particle size composition in ecoforest polygons for the entire area of managed forests in the province of Quebec, Canada. We compiled archived laboratory analyses of 29,570 mineral soil samples (17,901 sites) and a set of 33 covariates, including 22 variables related to climate, five related to soil characteristics, three to spatial position or spatial context, two to relief and topography, and one to vegetation. After five repeats of 5-fold cross-validation, results show that models that include two functionally independent values regarding particle size composition explain 60%, 34%, and 78% of the variance in sand, silt and clay fractions, respectively, with mean absolute errors ranging from 4.0% for the clay fraction to 9.5% for the sand fraction. The most important model variables are those observed in the field and those interpreted from aerial photography regarding soil characteristics, followed by those regarding elevation and climate. Our results compare favorably with those of previous soil texture mapping studies for the same territory, in which particle size composition was modeled mainly from rasterized climatic and topographic covariates. The map we provide should meet the needs of provincial forest managers, as it is compatible with the ecoforest map that constitutes the basis of information for forest management in Quebec, Canada.

Obfuscated Tor Traffic Identification Based on Sliding Window

Tor is an anonymous communication network used to hide the identities of both parties in communication. Apart from those who want to browse the web anonymously using Tor for a benign purpose, criminals can use Tor for criminal activities. It is recognized that Tor is easily intercepted by the censorship mechanism, so it uses a series of obfuscation mechanisms to avoid censorship, such as Meek, Format-Transforming Encryption (FTE), and Obfs4. In order to detect Tor traffic, we collect three kinds of obfuscated Tor traffic and then use a sliding window to extract 12 features from the stream according to the five-tuple, including the packet length, packet arrival time interval, and the proportion of the number of bytes sent and received. And finally, we use XGBoost, Random Forest, and other machine learning algorithms to identify obfuscated Tor traffic and its types. Our work provides a feasible method for countering obfuscated Tor network, which can identify the three kinds of obfuscated Tor traffic and achieve about 99% precision rate and recall rate.

Machine learning prediction of COVID-19 mortality in cancer patients.

1558 Background: COVID-19 is a challenge for clinical decision-making in cancer patients and the allocation of healthcare resources. An accurate prognosis prediction to effectively triage patients is needed, especially in the community oncology practice. Methods:Nationwide cohort from Oncoclínicas Brazil was used to validate previously developed multivariable logistic regression (mLR) model (Ferrari et al, JCO GO 2021) and to construct a machine learning Random Forest (RF) algorithm as predictor of 30-day mortality after SARS-CoV-2 detection by RT-PCR in cancer patients diagnosed in an outpatient setting. To find the most important baseline clinical determinants of early COVID-19-related death via Gini index, a RF with 100,000 trees was trained in 75% of the dataset, and the performance was assessed in the remaining 25%. We then compared the accuracy of different models in terms of sensitivity, specificity and area under the receiver operating characteristics curves (AUC). Results:From March to December 2020, 533 patients with COVID-19 were prospectively registered in the database. Median age was 60 years (19-93) and 67% were female. Most frequent cancers were breast in 34%, hematological in 16%, and gastrointestinal in 15%. Comorbidities were common (52%), as was current/former smoking history (17%). Most patients were on active systemic therapy or radiotherapy (84%) in the advanced or metastatic disease setting (55%). The overall mortality rate was 15% (CI95% 12%-18%). We validated the original mLR model trained in the first 198 patients: management in a non-curative setting (odds ratio [OR] 3.7), age ≥ 60 years (OR 2.3), and current/former smoking (OR 1.9) were significant predictors of death in the expanded cohort. Presence of comorbidities (OR 1.9) also defined poor outcome in the updated mLR model, which yielded low sensitivity (74%), specificity (68%) and AUC (0.78). With RF modeling, the most significant predictors of 30-day death after COVID-19 (in decreasing order) were older age, treatment of advanced or metastatic disease, tumor type (respiratory tract, brain and unknown primary cancers had higher mortality), COVID-related symptom burden at baseline evaluation and treatment regimen (immunotherapy combinations had higher mortality). The RF model demonstrated high sensitivity (89%), specificity (88%) and AUC (0.96). Conclusions:The results highlight the possibility that machine learning algorithms are able to predict early mortality after COVID-19 in cancer patients with high accuracy. The proposed prediction model may be helpful in the prompt identification of high-risk patients based on clinical features alone, without having to wait for the results of additional tests such as laboratory or radiologic studies. It can also help prioritize medical resources and redefine vaccination strategies. A web-based mortality risk calculator will be created for clinical decision support.

Predictive Modeling Framework for Diabetes Classification Using Big Data Tools and Machine Learning

Diabetics is now a days a common and most threatening disease irrespective of age, gender and becoming a human threat. As the Internet of Things (IoT) environment is growing rapidly in health sector and continuously gathering the data from smart health care which directly reflects the growth of the big data. Predictive modeling helps doctors and physicians to the identify the growth of diabetics in patients from early ages and create an alarm such to make the patient more attentive towards the diabetics. Based on the previous approaches on diabetics prediction over big data related diabetic prediction yields in better understanding from the patient perspective. The approach for the proposed system is much wider in term of predicting the diabetic model with enough feature variables explaining the patient historical data and diet habits. The Frame work has been carried based on extensive machine learning methods in association for processing of the data over spark RDD. The Random Forest and Ada Boost algorithm showed us a prominent values in terms of predicting the results.

Automatic brain tumor segmentation for a computer‐aided diagnosis system

AbstractBrain structure segmentation, including tumors, in medical imaging has become a necessity to help neurologists correctly diagnose patients' conditions. The complexity of these structures requires the implementation of automatic segmentation methods, often developed by magnetic resonance imaging. This study aims to design and implement an automatic system for detecting and localizing tumor regions by combining three different methods. Firstly, the region of interest, that is, the pixels belonging to the tumor, is detected using Random Forest's algorithm, while the rest of the pixels of the image are considered to belong to the background. Thus, the tumor and background seeds are obtained, automatically, for segmentation using the Graph Cut method. This segmentation allows to obtain the initial contour, for the level set (LVS) segmentation, which refine the previous segmentation. The proposed method was validated on the Multimodal Brain Tumor Segmentation Challenge (BRATS) database (http://braintumorsegmentation.org; 2015).

Factors affecting the performance of brain arteriovenous malformation rupture prediction models

BackgroundIn many cases, both the rupture rate of cerebral arteriovenous malformation (bAVM) in patients and the risk of endovascular or surgical treatment (when radiosurgery is not appropriate) are not low, it is important to assess the risk of rupture more cautiously before treatment. Based on the current high-risk predictors and clinical data, different sample sizes, sampling times and algorithms were used to build prediction models for the risk of hemorrhage in bAVM, and the accuracy and stability of the models were investigated. Our purpose was to remind researchers that there may be some pitfalls in developing similar prediction models.MethodsThe clinical data of 353 patients with bAVMs were collected. During the creation of prediction models for bAVM rupture, we changed the ratio of the training dataset to the test dataset, increased the number of sampling times, and built models for predicting bAVM rupture by the logistic regression (LR) algorithm and random forest (RF) algorithm. The area under the curve (AUC) was used to evaluate the predictive performances of those models.ResultsThe performances of the prediction models built by both algorithms were not ideal (AUCs: 0.7 or less). The AUCs from the models built by the LR algorithm with different sample sizes were better than those built by the RF algorithm (0.70 vs 0.68, p < 0.001). The standard deviations (SDs) of the AUCs from both prediction models with different sample sizes displayed wide ranges (max range > 0.1).ConclusionsBased on the current risk predictors, it may be difficult to build a stable and accurate prediction model for the hemorrhagic risk of bAVMs. Compared with sample size and algorithms, meaningful predictors are more important in establishing an accurate and stable prediction model.

Screening for Core Genes Related to Pathogenesis of Alzheimer's Disease.

Alzheimer’s disease (AD), a nervous system disease, lacks effective therapies at present. RNA expression is the basic way to regulate life activities, and identifying related characteristics in AD patients may aid the exploration of AD pathogenesis and treatment. This study developed a classifier that could accurately classify AD patients and healthy people, and then obtained 3 core genes that may be related to the pathogenesis of AD. To this end, RNA expression data of the middle temporal gyrus of AD patients were firstly downloaded from GEO database, and the data were then normalized using limma package following a supplementation of missing data by k-Nearest Neighbor (KNN) algorithm. Afterwards, the top 500 genes of the most feature importance were obtained through Max-Relevance and Min-Redundancy (mRMR) analysis, and based on these genes, a series of AD classifiers were constructed through Support Vector Machine (SVM), Random Forest (RF), and KNN algorithms. Then, the KNN classifier with the highest Matthews correlation coefficient (MCC) value composed of 14 genes in incremental feature selection (IFS) analysis was identified as the best AD classifier. As analyzed, the 14 genes played a pivotal role in determination of AD and may be core genes associated with the pathogenesis of AD. Finally, protein-protein interaction (PPI) network and Random Walk with Restart (RWR) analysis were applied to obtain core gene-associated genes, and key pathways related to AD were further analyzed. Overall, this study contributed to a deeper understanding of AD pathogenesis and provided theoretical guidance for related research and experiments.

Research on Camouflage Recognition in Simulated Operational Environment Based on Hyperspectral Imaging Technology

Hyperspectral imaging technology can obtain the spatial information and spectral information of the simulated operational background and its camouflage materials at the same time and identify and classify them according to their differences. In this paper, we collected the hyperspectral images (400–1000 nm) of the desert background, jungle background, desert camouflage netting, jungle camouflage netting, and jungle camouflage clothing through the hyperspectral imaging system, and the samples were preprocessed by denoising and black-and-white correction. Then, we analysed the region of interest (ROI) of the training samples by principal component analysis (PCA). After the pixels in the region of interest and their surrounding areas were averaged, 60% of the data was used as the training samples, and the remaining 40% was used as the test samples. According to their similarities and differences between them and referenced spectrum, the models of classification were established by combining the Naive Bayes (NB) algorithm, K-nearest neighbour (KNN) algorithm, random forest (RF) algorithm, and support vector machine (SVM) algorithm. The results show that among the four models, SVM model has the highest accuracy of classification and the recognition rate of jungle camouflage clothing is the highest. This study verifies the scientific and feasibility of hyperspectral imaging technology for camouflage identification and classification in a simulated operational environment, which has some practical significance.

A new spectral index for the quantitative identification of yellow rust using fungal spore information

ABSTRACT Yellow rust (Puccinia striiformis f. sp. Tritici) is a frequently occurring fungal disease of winter wheat (Triticum aestivum L.). During yellow rust infestation, fungal spores appear on the surface of the leaves as yellow and narrow stripes parallel to the leaf veins. We analyzed the effect of the fungal spores on the spectra of the diseased leaves to find a band sensitive to yellow rust and established a new vegetation index called the yellow rust spore index (YRSI). The estimation accuracy and stability were evaluated using two years of leaf spectral data, and the results were compared with eight indices commonly used for yellow rust detection. The results showed that the use of the YRSI ranked first for estimating the disease ratio for the 2017 spectral data (R2 = 0.710, RMSE = 0.097) and outperformed the published indices (R2 = 0.587, RMSE = 0.120) for the validation using the 2002 spectral data. The random forest (RF), k-nearest neighbor (KNN), and support vector machine (SVM) algorithms were used to test the discrimination ability of the YRSI and the eight commonly used indices using a mixed dataset of yellow-rust-infested, healthy, and aphid–infested wheat spectral data. The YRSI provided the best performance.

Identifying Top Predictors of Change in Noncalcified Coronary Burden in Psoriasis by Machine Learning Over 1-Year

Background: Psoriasis is associated with accelerated non-calcified coronary burden (NCB) by coronary computed tomography angiography (CCTA). Machine learning (ML) algorithms have been shown to effectively identify cardiometabolic variables with NCB in cross-sectional analysis. Objective: To use ML methods to characterize important predictors of change in NCB by CCTA in psoriasis over 1-year of observation. Methods: The analysis included 182 consecutive patients with 80 available variables from the Psoriasis Atherosclerosis Cardiometabolic Initiative, a prospective, observational cohort study at baseline and 1-year using the random forest regression algorithm. NCB was assessed at baseline and 1-year from CCTA. Results: Using ML, we identified variables of high importance in the context of predicting changes in NCB. For the cohort that improved NCB (n = 102), top baseline variables were cholesterol (total and HDL), white blood cell count, psoriasis area severity index score, and diastolic blood pressure. Top predictors of 1-year change were change in visceral adiposity, white blood cell count, total cholesterol, c-reactive protein, and absolute lymphocyte count. For the cohort that worsened NCB (n = 80), the top baseline variables were HDL cholesterol related including apolipoprotein A1, basophil count, and psoriasis area severity index score, and top predictors of 1-year change were change in apoA, apoB, and systolic blood pressure. Conclusion: ML methods ranked predictors of progression and regression of NCB in psoriasis over 1 year providing strong evidence to focus on treating LDL, blood pressure, and obesity; as well as the importance of controlling cutaneous disease in psoriasis.

Predicting classifications in marine biomonitoring with supervised machine learning: how much data is required?

Marine coastal ecosystems offer numerous ecosystem services and are therefore subject to a variety of stressors from anthropogenic activities. Environmental biomonitoring programs for effective management and conservation of coastal marine ecosystems are therefore crucial. Traditional monitoring has been based on macrofauna indices which are laborious and require expert knowledge. Recently, eDNA metabarcoding has become increasingly popular as it does not involve macrofauna species identification and is therefore cost and time inexpensive. Studies have shown that ecosystem monitoring based on eDNA metabarcoding is feasible and random forest (RF) algorithms can predict various biological indices, and therefore ecosystem health. To propose adequate designs for future eDNA metabarcoding-based marine coastal monitoring surveys, the aim of the study is to find out (1) What is the lower limit of reads for accurate RF predictions in coastal marine monitoring using microbial communities? (2) Is this limit the same for different monitoring targets? To achieve this goal, we exploited four different Illumina amplicon datasets obtained from bacterial communities in different costal environments. From these datasets, we predicted different objectives relevant for biomonitoring. For each dataset, those corresponding prediction objectives (labels) were predicted using amplicon sequence variants (ASVs) as features. After construction of RF models using all available sequences of a dataset (full model, serving as benchmark for targeted prediction accuracy), we then successively down-sampled each dataset to lower sequence numbers. Prediction accuracies of the reduced models were then compared to the accuracies of the full models to assess the minimum number of features to obtain the targeted prediction accuracy. Our results show that there is no general answer to question (1) and that (2) the limit varies between different monitoring targets. We have identified the most informative criteria that are relevant to assess the sequencing depth required to predict a biomonitoring category using RF. This may guide future study designs and may help to estimate and control costs in applied routine DNA-based biomonitoring using RF to predict the biomonitoring target. In our contribution we will elucidate and discuss these criteria.

Analysis of protein determinants of host-specific infection properties of polyomaviruses using machine learning.

The large tumor antigen (LT-Ag) and major capsid protein VP1 are known to play important roles in determining the host-specific infection properties of polyomaviruses (PyVs). The objective of this study was to investigate the physicochemical properties of amino acids of LT-Ag and VP1 that have important effects on host specificity, as well as classification techniques used to predict PyV hosts. We collected and used reference sequences of 86 viral species for analysis. Based on the clustering pattern of the reconstructed phylogenetic tree, the dataset was divided into three groups: mammalian, avian, and fish. We then used random forest (RF), naïve Bayes (NB), and k-nearest neighbors (kNN) algorithms for host classification. Among the three algorithms, classification accuracy using kNN was highest in both LT-Ag (ACC = 98.83) and VP1 (ACC = 96.51). The amino acid physicochemical property most strongly correlated with host classification was charge, followed by solvent accessibility, polarity, and hydrophobicity in LT-Ag. However, in VP1, amino acid composition showed the highest correlation with host classification, followed by charge, normalized van der Waals volume, and solvent accessibility. The results of the present study suggest the possibility of determining or predicting the host range and infection properties of PyVs at the molecular level by identifying the host species of active and emerging PyVs that exhibit different infection properties among diverse host species. Structural and biochemical differences of LT-Ag and VP1 proteins in host species that reflect these amino acid properties can be considered primary factors that determine the host specificity of PyV.

Spatial variation in lymphatic filariasis risk factors of hotspot zones in Ghana

BackgroundLymphatic Filariasis (LF), a parasitic nematode infection, poses a huge economic burden to affected countries. LF endemicity is localized and its prevalence is spatially heterogeneous. In Ghana, there exists differences in LF prevalence and multiplicity of symptoms in the country’s northern and southern parts. Species distribution models (SDMs) have been utilized to explore the suite of risk factors that influence the transmission of LF in these geographically distinct regions.MethodsPresence-absence records of microfilaria (mf) cases were stratified into northern and southern zones and used to run SDMs, while climate, socioeconomic, and land cover variables provided explanatory information. Generalized Linear Model (GLM), Generalized Boosted Model (GBM), Artificial Neural Network (ANN), Surface Range Envelope (SRE), Multivariate Adaptive Regression Splines (MARS), and Random Forests (RF) algorithms were run for both study zones and also for the entire country for comparison.ResultsBest model quality was obtained with RF and GBM algorithms with the highest Area under the Curve (AUC) of 0.98 and 0.95, respectively. The models predicted high suitable environments for LF transmission in the short grass savanna (northern) and coastal (southern) areas of Ghana. Mainly, land cover and socioeconomic variables such as proximity to inland water bodies and population density uniquely influenced LF transmission in the south. At the same time, poor housing was a distinctive risk factor in the north. Precipitation, temperature, slope, and poverty were common risk factors but with subtle variations in response values, which were confirmed by the countrywide model.ConclusionsThis study has demonstrated that different variable combinations influence the occurrence of lymphatic filariasis in northern and southern Ghana. Thus, an understanding of the geographic distinctness in risk factors is required to inform on the development of area-specific transmission control systems towards LF elimination in Ghana and internationally.

A machine learning-based predictor for the identification of the recurrence of patients with gastric cancer after operation

To explore the predictive performance of machine learning on the recurrence of patients with gastric cancer after the operation. The available data is divided into two parts. In particular, the first part is used as a training set (such as 80% of the original data), and the second part is used as a test set (the remaining 20% of the data). And we use fivefold cross-validation. The weight of recurrence factors shows the top four factors are BMI, Operation time, WGT and age in order. In training group:among the 5 machine learning models, the accuracy of gbm was 0.891, followed by gbm algorithm was 0.876; The AUC values of the five machine learning algorithms are from high to low as forest (0.962), gbm (0.922), GradientBoosting (0.898), DecisionTree (0.790) and Logistic (0.748). And the precision of the forest is the highest 0.957, followed by the GradientBoosting algorithm (0.878). At the same time, in the test group is as follows: the highest accuracy of Logistic was 0.801, followed by forest algorithm and gbm; the AUC values of the five algorithms are forest (0.795), GradientBoosting (0.774), DecisionTree (0.773), Logistic (0.771) and gbm (0.771), from high to low. Among the five machine learning algorithms, the highest precision rate of Logistic is 1.000, followed by the gbm (0.487). Machine learning can predict the recurrence of gastric cancer patients after an operation. Besides, the first four factors affecting postoperative recurrence of gastric cancer were BMI, Operation time, WGT and age.