Genetic Programming Method Research Articles

Refractive Index Formula of Blood as a Function of Temperature and Concentration.

The analysis of blood parameters is main procedure for defining the patient's condition. The refractive index of blood was calculated using the experimental data for a medical application and diagnostic purposes. There is small change in the refractive indices with changing with the personal conditions, illness, parasitization, temperature and others. Theoretical simulations of the refractive index of blood are difficult, and it is unpredictable in different conditions. We proposed a new formula for the refractive index of blood as a function of wavelength, concentration, and temperature by using the genetic programming method. Input parameters were temperature (°C), concentration (g/L), and wavelength (nm). The refractive index values of blood were the output parameters. A total of 492 training and testing sets were selected in the spectral range of 436 to 1550 nm, in the temperature range of 20-45 °C and the concentration of 0-200 g/L HbC. The model proposes the refractive index formula blood for all the input parameters given in the range without need of extra parameters. The results are good agreement with experimental measurements in the literature and compared to Sellmeier equation.

Sarima-arch versus genetic programming in stock price prediction

In financial time series, one of the most challenging problems is predicting stock prices since the data generally exhibit deviation from the assumptions of stationary and homoscedasticity. For homogenous non-stationary time series, the Autoregressive Integrated Moving Average (ARIMA) model is the most commonly used linear class including some transformation such as differencing and variance stabilizing process. However, stock market data is often nonlinear, which indicates that more advanced methods are necessary. Genetic Programming (GP) is one of the evolutionary computational methods that could capture both linear and nonlinear patterns in time series data. The present study aims to build a machine learning tool using GP for prediction The Istanbul Stock Exchange National 100 (XU100) index and compare the obtained results with conventional seasonal ARIMA(SARIMA) and ARCH models. In order to achieve this goal, it was first modeled with the SARIMA model after appropriate transfor- mations were made to the stock price series and the diagnostic control result showed that the residual of the SARIMA model have the heteroscedasticity problem. Then, the ARCH model was applied to SARIMA residuals to eliminate this effect and an integrated SARIMA-ARCH model is obtained. Since it is possible and capable to model nonlinear and non-stationary time series using GP without any pre-assumptions, we proposed GP to predict the stock price series. The function set of GP consists of not only arithmetic but also trigonometric functions. To the best of our knowledge, this study is the first to predict XU100 stock price data using GP. In this experiment, the data set consists of the daily closing prices of the XU100 index over 775 days from the beginning of 2017 until the end of January 2020. The experimental results obtained show that the accuracy metrics used in the study are lower in the proposed GP model compared to other models. These results reveal that the GP method provides better predictive results for the financial time series data of the XU100 index than traditional methods.

On the evaluation of solubility of hydrogen sulfide in ionic liquids using advanced committee machine intelligent systems

Ionic Liquids (ILs) are increasingly emerging as new innovating green solvents with great importance from academic, industrial, and environmental perspectives. This surge of interest in considering ILs in various applications is owed to their attractive properties. Involvements in the gas sweetening and the reduction of the amounts of sour and acid gasses are among the most promising applications of ILs. In this study, new advanced committee machine intelligent systems (CMIS) were introduced for predicting the solubility of hydrogen sulfide (H2S) in various ILs. The implemented CMIS models were gained by linking robust data-driven techniques, namely multilayer perceptron (MLP) and cascaded forward neural network (CFNN) beneath rigorous schemes using group method of data handling (GMDH) and genetic programming (GP). The proposed paradigms were developed using an extensive database encompassing 1243 measurements of H2S solubility in 33 ILs. The performed comprehensive error investigation revealed that the newly implemented paradigms yielded very satisfactory prediction performance. Besides, it was found that CMIS-GP provided more accurate estimations of H2S solubility in ILs compared with both the other intelligent models and the best-prior paradigms. In this regard, the developed CMIS-GP exhibited overall average absolute relative deviation (AARD) and coefficient of determination (R2) values of 2.3767% and 0.9990, respectively. Lastly, the trend analyses demonstrated that the tendencies of CMIS-GP predictions were in excellent accordance with the real variations of H2S solubility in ILs with respect to pressure and temperature.

Robust and General Model to Forecast the Heat Transfer Coefficient for Flow Condensation in Multi Port Mini/Micro-Channels

A general correlation for predicting the two-phase heat transfer coefficient (HTC) during condensation inside multi-port mini/micro-channels was presented. The model was obtained by correlating the two-phase multiplier, [[EQUATION]] with affecting parameters using the genetic programming (GP) method. An extensive database containing 3503 experimental data samples was gathered from 21 different sources, including a broad range of operating parameters. The newly obtained correlation fits the broad range of measured data analyzed with an average absolute relative deviation (AARD) of 16.87% and estimates 84.73% of analyzed data points with a relative error of less than 30%. Evaluation of previous correlations was also done using the same database. They showed the AARD values ranging from 36.94% to 191.19%. However, the GP model provides more accurate results, AARD lower than 17%, by considering the surface tension effects. Finally, the effect of various operating parameters on the HTC was studied using the proposed correlation.

Applying genetic programming in estimation of frost layer thickness on horizontal and vertical plates at ultra-low temperature

Abstract In this study, the intelligent method of genetic programming (GP) was used for developing predictive models for estimating the frost layer thickness under natural and forced convection on ultra-low temperature surfaces. The affecting dimensionless parameters were used as GP input variables, and realistic empirical correlations were developed for estimating the frost thickness under different conditions. The coefficient of determination of 0.9731, 0.9812, and 0.9906, and average absolute relative error of 6.52%, 11.65%, and 2.87%, were obtained by the developed models for natural convection on vertical plates, natural convection on horizontal plates, and forced convection on horizontal plates, respectively. The physical trends of the developed models were evaluated by comparing the model predictions with the experimental data for different operating conditions, and reasonable agreements were obtained. The same experimental database was also compared to some existing correlations for ordinary-low temperature surfaces, but they failed to provide reasonable estimates for the data.

Genetic programming for development of cost-sensitive classifiers for binary high-dimensional unbalanced classification

Genetic programming (GP) has the built-in ability for feature selection when developing classifiers for classification with high-dimensional data. However, due to the problem of class imbalance, the developed classifiers by GP are prone to be biased towards the majority class. Cost-sensitive learning has shown to be effective in addressing the problem of class imbalance. In cost-sensitive learning, cost matrices are often manually designed and then considered by classification algorithms to treat different mistakes differently. However, in many real-world applications, cost matrices are unknown because of the limited domain knowledge in complex situations. Therefore, in this paper, we propose a novel GP method to develop cost-sensitive classifiers, where a cost matrix is automatically learned, instead of requiring it from domain experts. The proposed method is examined and compared with existing methods on ten high-dimensional unbalanced datasets. Experimental results show that the proposed method outperforms the compared GP methods in most cases.

Forecasting Models of the Coronavirus (COVID-19) Cumulative Confirmed Cases Using a Hybrid Genetic Programming Method

The coronavirus disease 2019 (COVID-19) diffusion process, starting in China, caused more than 4600 lives until June 2020 and became a major threat to global public health systems. In Greece, the phenomenon started on February 2020 and it is still being developed. This paper presents the implementation of a hybrid Genetic Programming (hGP) method in finding fitting models of the Coronavirus (COVID 19) for the cumulative confirmed cases in China for the first saturation level until May 2020 and it proposes forecasting models for Greece before summer tourist season. The specific hGP method encapsulates the use of some well-known diffusion models for forecasting purposes, epidemiological models and produces time dependent models with high performance statistical indices. A retrospective study confirmed the excellent forecasting performance of the method until 3 June 2020.

Forecasting Models of the Coronavirus (COVID-19) Cumulative Confirmed Cases Using a Hybrid Genetic Programming Method

The coronavirus disease 2019 (COVID-19) diffusion process, starting in China, caused more than 4600 lives until June 2020 and became a major threat to global public health systems. In Greece, the phenomenon started on February 2020 and it is still being developed. This paper presents the implementation of a hybrid Genetic Programming (hGP) method in finding fitting models of the Coronavirus (COVID 19) for the cumulative confirmed cases in China for the first saturation level until May 2020 and it proposes forecasting models for Greece before summer tourist season. The specific hGP method encapsulates the use of some well-known diffusion models for forecasting purposes, epidemiological models and produces time dependent models with high performance statistical indices. A retrospective study confirmed the excellent forecasting performance of the method until 3 June 2020.

Genetic Programming Method for Satellite System Topology and Parameter Optimization

In this paper, a genetic programming method for satellite system design is proposed to simultaneously optimize the topology and parameters of a satellite system. Firstly, the representation of satellite system design is defined according to the tree structure. The genetic programming method is designed based on that representation. Secondly, according to the tree structure of different satellite schemes, different multiscale satellite models are established, in which various physical fields couple together. Then, an evaluation system is also proposed to test the performances of different satellite schemes. Finally, the application to the design of an earth observation satellite demonstrates the effectiveness of the proposed method.

SVM &Ga-clustering Based Feature Selection Approach for Breast Cancer Detection

Mortality leading among women in developed countries is breast cancer. Breast cancer is women's second most prominent cause of cancer mortality worldwide. In recent decades, women's high prevalence of breast cancer has risen dramatically. This paper discussed several data analysis methods used to detect breast cancer early. Breast cancer diagnosis distinguishes benign and malignant breast lumps. Using data processing tools, we tackled this disease analysis. Data mining is an important step of library discovery where intelligent methods are used to detect patterns. Several clinical breast cancer studies were conducted using soft computing and machine learning techniques. Sometimes their algorithms are easier, easier, or more comprehensive than others. This research is focused on genetic programming and machine learning algorithms to reliably identify benign and malignant breast cancer. This study aimed to optimise the testing algorithm. We used genetic programming methods to choose classification machines' best features and parameter values. Data mining is an important step of library discovery where intelligent methods are used to detect patterns. We are analysing data accessible from the U.C.I. deep-learning data set in Wisconsin. In this experiment, we equate four Weka clustering strategies with genetic clustering. A comparison of results reveals that sequential minimal optimization (S.M.O.) is better than I.B.K. and B.F. Tree processes, i.e. 97.71%.

Hybrid soft computational approaches for modeling the maximum ultimate bond strength between the corroded steel reinforcement and surrounding concrete

The capacity efficiency of load carrying with the accurate serviceability performances of reinforced concrete (RC) structure is an important aspect, which is mainly dependent on the values of the ultimate bond strength between the corroded steel reinforcements and the surrounding concrete. Therefore, the precise determination of the ultimate bond strength degradation is of paramount importance for maintaining the safety levels of RC structures affected by corrosion. In this regard, hybrid intelligence and machine learning techniques are proposed to build a new framework to predict the ultimate bond strength in between the corroded steel reinforcements and the surrounding concrete. The proposed computational techniques include the multilayer perceptron (MLP), the radial basis function neural network and the genetic expression programming methods. In addition to that, the Levenberg–Marquardt (LM) deterministic approach and two meta-heuristic optimization approaches, namely the artificial bee colony algorithm and the particle swarm optimization algorithm, are employed in order to guarantee an optimum selection of the hyper-parameters of the proposed techniques. The latter were implemented based on an experimental published database consisted of 218 experimental tests, which cover various factors related to the ultimate bond strength, such as compressive strength of the concrete, concrete cover, the type steel, steel bar diameter, length of the bond and the level of corrosion. Based on their performance evaluation through several statistical assessment tools, the proposed models were shown to predict the ultimate bond strength accurately; outperforming the existing hybrid artificial intelligence models developed based on the same collected database. More precisely, the MLP-LM model was, by far, the best model with a determination coefficient (R2) as high as 0.97 and 0.96 for the training and the overall data, respectively.

A hyper-heuristic based ensemble genetic programming approach for stochastic resource constrained project scheduling problem

In project scheduling studies, to the best of our knowledge, the hyper-heuristic collaborative scheduling is first-time applied to project scheduling with random activity durations. A hyper-heuristic based ensemble genetic programming (HH-EGP) method is proposed for solving stochastic resource constrained project scheduling problem (SRCPSP) by evolving an ensemble of priority rules (PRs). The proposed approach features with (1) integrating the critical path method into the resource-based policy class to generate schedules; (2) improving the existing single hyper-heuristic project scheduling research to construct a suitable solution space for solving SRCPSP; and (3) bettering genetic evolution of each subpopulation from a decision ensemble with three different local searches in corporation with discriminant mutation and discriminant population renewal. In addition, a sequence voting mechanism is designed to deal with collaborative decision-making in the scheduling process for SRCPSP. The benchmark PSPLIB is performed to verify the advantage of the HH-EGP over heuristics, meta-heuristics and the single hyper-heuristic approaches.

Modeling and predicting chloride diffusion in recycled aggregate concrete

Application of recycled aggregate concrete (RAC) in engineering practice today remains relatively limited. One of the main reasons may be that the material’s durability has not been comprehensively understood. No sufficiently accurate formulas are available for predicting its resistance to chloride infiltration. This study was therefore designed to investigate the commonest factors influencing chloride penetration in RAC using mesoscale finite element models. The variables of interest were the geometric shape of coarse aggregate pieces, their location distribution, the volume content of recycled material, the relative strength of the old to new mortar, the adhering content of old mortar, the bonding property of interfacial transition zones (ITZs) and the mixing method used. After performing a series of numerical simulations, a genetic programming (GP) method was lastly adopted to establish an explicit expression for correlating the RAC’s effective chloride diffusivity with the identified key factors. Numerical results indicate that the RAC’s diffusion coefficient was negligibly influenced by the aggregate shape or the old ITZ property, and commonly grows with increasing water-to-cement ratio, the amount of old mortar, the new ITZ’s diffusivity as well with the replacing content of recycled aggregates. Equivalent mortar volume method can efficiently decrease the material’s chloride diffusivity, especially at low water-to-cement ratios in the attached mortar. Finally, the expression provided by the GP method can adequately predict all these trends and is very convenient for investigating the RAC’s chloride diffusion performance.

Prediction and pareto-based multi-objective optimization of moisture and fatigue damages of asphalt mixtures modified with nano hydrated lime

Moisture sensitivity can lead to damages due to adhesion and cohesion in asphalt mixtures that cause serious damages to the pavement. Adhesion between bitumen and aggregate is a basic prerequisite for the proper performance of asphalt mixtures. The role of nano hydrated lime (NHL) as an anti-stripping agent has been extensively studied in asphalt pavements. In this study, in addition to investigating the effect of NHL as bitumen modifier on moisture and fatigue damages, and bitumen-aggregate adhesion and bitumen membrane cohesion using surface free energy (SFE) concept, the prediction model to achieve optimal additive percentage with two objectives of moisture sensitivity and fatigue failure by multivariate regression (MVR), group method of data handling (GMDH) and genetic programming (GP) were provided. By the use of the modified non-dominated sorting genetic algorithm II (NSGA-II) approach, Pareto fronts were then plotted. In addition, boiling water testing was applied to examine the sensitivity of mixtures to moisture. To simulate environmental conditions, asphalt mixtures were placed in freeze–thaw (F-T) cycles of 1, 3 and 5. 85–100 penetration grade base bitumen was modified with 0.5%, 1% and 1.5% NHL (by the weight of bitumen). The results indicated that by decreasing the number of yellow pixels in the image processing of modified mixtures with NHL, modification of bitumens decreased the moisture sensitivity of specimens. Using NHL increased the resistance of the specimens to moisture and fatigue damages. The amount of indirect tensile strength (ITS) and fatigue life of asphalt mixtures reduced by rising the number of F-T cycles. The highest increase in ITS and fatigue life was obvious in modified mixtures with 1.5% and 1% NHL, respectively. The results of calculating the surface free energy (SFE) components of the base and modified bitumens showed that using NHL enhanced the total SFE amount of bitumens. Also, using the NHL increased cohesion free energy (CFE), adhesion free energy (AFE) and debonding energy (DE) of modified bitumens and decreased the permeability of asphalt mixtures (PAMs) amount. Among the various methods, the GMDH method had the highest R2. So that the R2 value of GMDH for tensile strength ratio (TSR) and the ratio of fatigue life (NFR) was 93.9% and 86.9%, respectively. The two-objective optimization results indicated that 1.29% NHL was the best optimal amount to maximize the TSR and NFR values simultaneously.

A practical design of hash functions for IPv6 using multi-objective genetic programming

Hash functions are widely used in high-speed network traffic measurement. A hash function of high quality is supposed to meet the requirements of collision free and fast execution. Existing works have already developed methods to generate hash functions for IPv4 data, while IPv6 data with much longer addresses and different data characteristics may decline the effectiveness of those methods. In this paper, we present a practical design of hash functions for IPv6 measurement, based on the entropy analysis of IPv6 network data and an automated method of multi-objective genetic programming (GP). Considering our specific application of hash functions, we use three fitness functions as the optimization objectives, including active flow estimation, uniformity and seed avalanche effect, among which the active flow estimation is the main objective as the specific measurement task. In implementation of multi-objective GP, we adopted a strategy to limit the hash functions to shorter execution time than other hash functions by advanced experimental investigation. Experiments were conducted to construct hash functions for WIDE IPv6 network data. The results show that our generated hash functions have high usability on different evaluation criteria. It indicates that our generated hash functions are superior in active flow estimation and execution time and could compete with state of art hash functions in terms of uniformity and generating independent hash values for data structures like Bloom Filter.

Investigation of shear strength correlations and reliability assessments of sandwich structures by kriging method

Steel-concrete-steel (SCS) sandwich composite structure with corrugated-strip connectors (CSC) is the promising structure which is applied in offshore and building structures. The behavior prediction of shear connections is of major importance in SCS structures. The present study evaluated the existing shear strength correlations of SCS sandwich structures exploiting experimental data and Finite Element Analysis (FEA). The considered system is a double steel skin sandwich structure with CSC (DSCS). Due to the limitation of the literature regarding CSC development, some new correlations were proposed and studied relying on several FEA results through the Genetic Programming method. The accuracy of the estimated shear strength predicted by the existing and proposed equations was evaluated using the FEA data and push-out test results. The FE models were verified through experimental data. Moreover, the correlations were investigated based on reliability assessment due to the high importance of the reliability analysis of such structures. Given that high accuracy in estimating the shear strength fails to necessarily lead to acceptable results in structural reliability analysis, the reliability of the existing and proposed equations was evaluated using the Kriging model by considering experimental data. This meta-model could predict accurate values with a limited number of initial training samples.

A GP-based ensemble classification framework for time-changing streams of intrusion detection data

Intrusion detection tools have largely benefitted from the usage of supervised classification methods developed in the field of data mining. However, the data produced by modern system/network logs pose many problems, such as the streaming and non-stationary nature of such data, their volume and velocity, and the presence of imbalanced classes. Classifier ensembles look a valid solution for this scenario, owing to their flexibility and scalability. In particular, data-driven schemes for combining the predictions of multiple classifiers have been shown superior to traditional fixed aggregation criteria (e.g., predictions’ averaging and weighted voting). In intrusion detection settings, however, such schemes must be devised in an efficient way, since (part of) the ensemble may need to be re-trained frequently. A novel ensemble-based framework is proposed here for the online intrusion detection, where the ensemble is updated through an incremental stream-oriented learning scheme, correspondingly to the detection of concept drifts. Differently from mainstream ensemble-based approaches in the field, our proposal relies on deriving, though an efficient genetic programming (GP) method, an expressive kind of combiner function defined in terms of (non-trainable) aggregation functions. This approach is supported by a system architecture, which integrates different kinds of functionalities, ranging from the drift detection, to the induction and replacement of base classifiers, up to the distributed computation of GP-based combiners. Experiments on both artificial and real-life datasets confirmed the validity of the approach.

Metal-doped bioceramic nanopowders with tunable structural properties aimed at enhancing bone density: Rapid synthesis and modeling

Metal doped bioceramic nanopowders were prepared by solid-state mechanochemical reactions. Also, genetic programming (GP) and gene expression programming (GEP) models were developed to predict the structural features of the mechanosynthesized nanopowders aimed at developing an innovative solution to enhance bone mineral density. The substitution of Ca2+ with different ions in the apatite structure was confirmed from chemical analysis and structural assessment, where irregular changes in the lattice parameters and unit cell volume were observed due to the replacement of the Ca2+ bivalent cation with monovalent, bivalent or trivalent ions as well as the carbonate ions effects on the apatite lattice. It was found that the crystallite size and micro-strain of the substituted bioceramics were between ~11 and 98 nm and ~0.31–2.49%, respectively. From the functional group analysis, the intensity of the hydroxyl groups decreased as the dopant content increased. The electron microscopy images showed that both undoped and low-doped samples consist of spheroidal particles in the nano regime, whereas the high-doped specimens exhibited a high propensity to agglomerate. The results of cytotoxicity assays corroborated that appropriate ionic substitution can prevent the toxic effects of Li on Mus musculus fibroblast cells, and thus by increasing dopant concentration up to z = 0.25, cell viability of around 90% was observed. The results obtained from the modeling demonstrated that both GP and GEP methods are reliable in predicting the structural properties of the synthetic metal-doped bioceramic nanopowders.

Rademacher Complexity for Enhancing the Generalization of Genetic Programming for Symbolic Regression.

Model complexity has a close relationship with the generalization ability and the interpretability of the learned models. Simple models are more likely to generalize well and easy to interpret. However, too much emphasis on minimizing complexity can prevent the discovery of more complex yet more accurate solutions. Genetic programming (GP) has a trend of generating overcomplex models that are difficult to interpret while not being able to generalize well. This work proposes a novel complexity measure based on the Rademacher complexity for GP for symbolic regression. The complexity of an evolved model is measured by the maximum correlation between the model and the Rademacher variables on the selected training instances. Taking minimizing the training error and the Rademacher complexity of the models as the two objectives, the proposed GP method has shown to be much superior to the standard GP on generalization performance. Compared with GP equipped with two state-of-the-art complexity measures, the proposed method still has a notable advance on generating a better front consisting of individuals with lower generalization errors and being simpler in the behavioral complexity. Further analyses reveal that compared with the state-of-the-art methods, the proposed GP method evolves models that are much closer to the target models in the model structure, and have better interpretability.

Prediction of Marshall Mix Design Parameters in Flexible Pavements Using Genetic Programming

The mix design of asphalt concrete is usually accomplished in the Iranian ministry of road and transportation according to the Marshall method. Marshall mix design parameters are a function of grading and properties of aggregates, amount and type of bitumen in asphalt mixtures. Therefore, in order to determine these parameters and the optimum bitumen content, many samples with different compounds and conditions must be manufactured and tested in the laboratory, a process that requires considerable time and cost. Accordingly, the necessity of using new and advanced methods for the design and quality control of asphalt mixtures is becoming more and more evident. Therefore, in this study, a genetic programming simulation method was employed to predict the Marshall mix design parameters of asphalt mixtures. Also, multiple linear regression models were adopted as the base model to evaluate the models presented by the genetic programming method. The models proposed here predict the Marshall mix design parameters based on parameters such as the index of aggregate particle shape and texture, the amount and viscosity of the bitumen. The results demonstrated that the proposed methods are more efficient than the costly laboratory method, and genetic programming models with minimal error (identified in this study with RMSE and MAE parameters) and correlation coefficients > 0.9 can predict relatively accurate Marshall mix design parameters.