China
Africa
Explore chapters and articles related to this topic
Knowledge Mining from Medical Images
Published in Wahiba Ben Abdessalem Karaa, Nilanjan Dey, Mining Multimedia Documents, 2017
Amira S. Ashour, Nilanjan Dey, Suresh Chandra Satapathy
Thus, data mining is known as a nontrivial extraction process of implied and impacted convenient information from the stored data in a database. It becomes the magic solution for KDD systems in various applications. The applied data mining methods to the medical data include unsupervised neural networks, support vector machine, apriori and FPGrowth, linear genetic programming, bayesian ying yang, association rule mining, naïve bayes and map (SOM), bayesian network, time series technique, clustering and classification, and decision tree algorithms [12–18].
Genetic Programming
Published in A Vasuki, Nature-Inspired Optimization Algorithms, 2020
There are different types of GP [6] such as linear genetic programming (LGP), traceless genetic programming (TGP), gene expression programming (GEP), multi-expression programming (MEP), Cartesian genetic programming (CGP), grammatical evolution (GE), genetic algorithm for deriving software (GADS), and fuzzy genetic programming (FGP).
Prophecy of Sediment Load Using Hybrid AI Approaches at Various Gauge Station in Mahanadi River Basin, India
Published in Sandeep Samantaray, Abinash Sahoo, Dillip K. Ghose, Watershed Management and Applications of AI, 2021
Sandeep Samantaray, Abinash Sahoo, Dillip K. Ghose
Kumar et al. (2016) employed ANN, RBFN, least-square SVM, classification and regression tree (CART), and M5 to predict SSC at River Kopili in India. Overall performance of models exhibited that all studied models simulated SSC of Kopili River basin adequately. Mustafa and Isa (2014) investigated the applicability of RBF and MLP techniques to predict SSC in Pari River, Perak, Malaysia. Results based on statistical parameters showed that performances of both RBF and MLP models were close to each other. Yet, RBFN revealed certain irregularity during time series data prediction. Memarian and Balasundram (2012) explored the potential of ANN, RBFN, and MLP models to predict SSC at Hulu Langat catchment, Malaysia. Findings from the study indicated that MLP showed a slightly better output than RBFN and ANN models to predict SSC. Malik et al. (2017) applied co-active neuro-fuzzy inference system (CANFIS), MLP, MLR, MNLR, and SRC methods to simulate daily SSC at Tekra gauge station on River Pranhita, Andhra Pradesh, India. Outcomes indicated supremacy of CANFIS model compared to other proposed models in simulating SSC for selected study area. Mohamadi et al. (2020) utilized hybrid neural network techniques integrating shark algorithm (SA) and FFA with ANFIS, MLP, and RBFN and also simple ANFIS, MLP, and RBFN to predict monthly evaporation of Mianeh and Yazd stations located in Iran. Based on evaluation criteria, results proved that developed ANFIS-SA hybrid model was considered as a powerful tool to predict evaporation. Kişi (2004) predicted and estimated SSC utilizing MLP at two gauging sites on River Tongue in Montana, USA, and compared the achieved results with RBF, MLR, and generalized regression NN (GRNN). Based on comparisons, it was found that MLP generally gave better SSC estimation than other NN methods. Olyaie et al. (2017) used MLP, RBF, SVM, and linear genetic programming (LGP) techniques to predict dissolved oxygen in River Delaware situated at Trenton, USA. Comparing estimation accuracies of proposed models demonstrated that SVM successfully developed most precise model for estimating DO followed by LGP and ANN models. Many researchers are now considering hybrid neural network to predict suspended sediment load (SSL) at different gauge station around the world (Samantaray and Ghose, 2019; Samantaray and Ghose, 2020; Samantaray et al., 2020b).
Improving Naive Bayes for Regression with Optimized Artificial Surrogate Data
Published in Applied Artificial Intelligence, 2020
As mentioned, we have chosen a direct representation in which variables being optimized map directly to attribute values in the artificial dataset. A consequence of this is that different particles (the equivalent of “genotypes” in evolutionary computing) may result in identical models (i.e., the equivalent of “phenotypes”) if the learning algorithm ignores the order of the examples in the data. For example, the datasets and , where is an example, produce the same NBR model. This observation is not necessarily a drawback of our approach. In fact, in many other approaches, multiple genotypes mapping onto a single phenotype is a deliberate feature as it enables neutral mutations. For example, linear genetic programming (LGP), proposed by (Brameier and Banzhaf 2007), evolves sequential programs. Since the order of any two instructions is often not dependent (e.g., initializing two different variables), the same issue arises: different programs map to the same behaviors. However, this does suggest that future research could investigate alternative particle/dataset mapping functions to determine if efficiencies are possible over the approach we use here.
Forecasting river basin yield using information of large-scale coupled atmospheric–oceanic circulation and local outgoing longwave radiation
Published in ISH Journal of Hydraulic Engineering, 2023
Satyawan D. Jagdale, Satishkumar S. Kashid, Ajay U. Chavadekar
The implementation of GP in this work is done using software tools viz. ‘Discipulus’ (Francone 1998) that is based on an extension of the originally envisaged GP called Linear Genetic Programming (LGP). The LGP evolves sequences of instructions from an imperative programming language or machine language. The LGP expresses instructions in a line-by-line mode. The term ‘linear’ in Linear Genetic Programming refers to the structure of the (imperative) program representation. It does not stand for functional genetic programs that are restricted to a linear list of nodes only. Genetic programs normally represent highly non-linear solutions in this meaning. (Brameier 2004).