Traditional Decision Tree Research Articles

An Experimental Comparison of Self-Adaptive Differential Evolution Algorithms to Induce Oblique Decision Trees

This study addresses the challenge of generating accurate and compact oblique decision trees using self-adaptive differential evolution algorithms. Although traditional decision tree induction methods create explainable models, they often fail to achieve optimal classification accuracy. To overcome these limitations, other strategies, such as those based on evolutionary computation, have been proposed in the literature. In particular, we evaluate the use of self-adaptive differential evolution variants to evolve a population of oblique decision trees encoded as real-valued vectors. Our proposal includes (1) an alternative initialization strategy that reduces redundant nodes and (2) a fitness function that penalizes excessive leaf nodes, promoting smaller and more accurate decision trees. We perform a comparative performance analysis of these differential evolution variants, showing that while they exhibit similar statistical behavior, the Single-Objective real-parameter optimization (jSO) method produces the most accurate oblique decision trees and is second best in compactness. The findings highlight the potential of self-adaptive differential evolution algorithms to improve the effectiveness of oblique decision trees in machine learning applications.

Approximating decision trees with priority hypotheses

This paper addresses the problem of creating decision trees for identifying hypotheses, also known as entities, in a setting where the cost of an action is dependent on the true hypothesis. Specifically, we consider the scenario where n hypotheses are divided into m groups based on their priority levels. Taking an action on a higher priority hypothesis incurs a higher cost. This is relevant to many real-world applications where cost-sensitive decisions need to be made. For example, in a medical diagnosis task, the goal is to take a series of actions (such as medical tests) to identify a cause. Each action in this process requires conducting a test on the patient and observing the outcome, which can take anywhere from a few minutes to several weeks depending on the test. In this case, the cost (the result of waiting for the outcome) is higher if the true hypothesis is more time-sensitive. For example, if the true hypothesis is toxic chemical exposure (as opposed to a chronic disease such as diabetes), a delay of a few minutes could significantly increase the patient's risk of mortality. We propose a group greedy algorithm to solve this problem. We demonstrate that under worst-case scenarios, our algorithm has an approximation ratio of O(mlog⁡n). Importantly, when m=1, meaning there is only one group of hypotheses, our result is consistent with the logarithmic approximation bound for the traditional optimal decision tree problem.

Predicting borrower default risk using support vector machine AI models

Precise prediction on the likelihood of borrower default is pivotal for credit institution and decision makers to mitigate the loss of capital and rationalize decision process. This article reviewed the Effects of Support Vector Machine (SVM) models with radial basis function (RBF) kernel in predicting the mortality rate of borrowers. By integrating with a dataset of approximately 100,000 borrowers profile harvested through historical loan performance, we set up the SVM model, and employed a feature-distribution method utilizing grid search and cross-validation technique to fine-tune the predictive model of SVM. Results indicated that the model accomplished an excellent performance with accuracy of 92%, precision of 89%, the recall and F1-score of 85% and 87%, respectively, alongside an Area Under the Curve -Receiver Operating Characteristic (AUC-ROC value of 0.95). It was evinced that the model performed substantially better than traditional logistic regression and decision trees in discriminating defaulter from non-defaulter. The outcome informs that an in-depth process should be implemented on data preprocessing, featureselection, and parameter tuning to achieve a robust predictive model for credit risk assessment. The article concludes the potentials of AI based on the resort to artificial technology in revolutionising the risk assessment scheme within the financial industry.

Research and Practice on Instructional Methods for Piano Improvization Based on Computer Technology

In the backdrop of evolving technological landscapes, this paper delves into the utilization of computer technology, specifically Artificial Neural Networks (ANN), in enhancing piano improvisation instruction. Recognizing the growing demand for innovative teaching methods, our study aims to evaluate the practical effectiveness of ANN-based teaching evaluation in real-world piano improvisation settings. Through rigorous testing, we found that ANN demonstrates remarkable precision and consistency in assessing piano improvisation skills. In comparison to the traditional decision tree algorithm, ANN excels in managing complex nonlinear relationships, providing more accurate and reliable scoring results. This integration of technology not only elevates students’ performance but also fosters their musical creativity and perception. Our findings suggest potential improvements in refining instructional methodologies and expanding the use of computer technology in piano teaching. This underscores the importance of personalized teaching, blended learning models, and technical proficiency training for educators. Overall, our research methodologies and findings significantly contribute to advancing the modernization and technological progress of music education, equipping piano instructors with cutting-edge teaching tools and strategies.

An efficient ranking-based data-driven model for ship inspection optimization

Maritime safety and environmental protection are fundamental considerations within the shipping industry. In this context, port state control (PSC) inspection is globally implemented by port authorities as a mechanism to enforce both maritime safety standards and environmental regulations. This study proposes an innovative optimization framework based on machine learning (ML) and operations research models for high-risk vessel selection, aiming to maximize the efficiency and effectiveness of PSC inspection. The essence of the optimization framework is to accurately rank all ships with respect to their risk levels predicted by ML models. The loss functions of the tailored ML models follow a “smart predict then optimize” (SPO) criterion named cumulative detected deficiency number (CDDN), which is motivated by the characteristics of the decision problem. This inventive measurement transforms the assessment of ranking accuracy to the area of the segmented histogram of the recognized deficiency number, which bypasses the computationally intensive training step of rankings and is easy to compute. Following this, three types of decision tree (DT) models are developed, which differ from each other in the varying integration levels of CDDN. Particularly, we rigorously prove that one integration method yields a tree structure identical to that of traditional DT models. The proposed models are validated and compared with the traditional DT model on different scales of instances from real inspection records at the Hong Kong port. The experiment results indicate that our tailored DT models improve the ship selection efficiency significantly when the decision is complex, i.e., when we need to optimize the selection of a small number of ships for inspection from a large number of foreign visiting ships. Moreover, we also extensively discuss when and why the SPO framework offers a superior decision to optimize vessel selection.

Stress Recognition in Code-Mixed Social Media Texts using Machine Learning

Stress, being a complex emotional state caused by a variety of multiple sources, has the potential for serious effects if left untreated. The primary goal of this research is to select and consider AI models that effectively recognize stress within the complicated domain of social media posts. The significance of this study is not only the categorization of stress but also the interpretation of the sophisticated methods that serve as the basis for these emotional responses. Among the traditional machine learning models, Random Forest, K-Nearest Neighbor, Logistic Regression, Decision Tree, and Support Vector Machine are used. The deep learning model’s LSTM, BiLSTM, and transformer-based models m-BERT, AL-BERT, XLM-RoBERTa, IndicBERT, and Distil-BERT were used. Of those models, LSTM proved to be the best-performing model, with an F1-score of 0.75.

Matching the building footprints of different vector spatial datasets at a similar scale based on one-class support vector machines

Automatic matching of multisource data is an important technique for achieving change detection, fusion and updating spatial data. However, most current learning methods for building footprint matching require a large number of samples, and labeling these samples is costly in terms of labor and time. Moreover, multisource building footprint data are complex and diverse leading to recognizing the different matching relationships is a hard task. Thus, this study proposes a learning-based method for recognizing multisource building footprints matching relationships by using a one-class support vector machine (OCSVM). The OCSVM was trained using only positive samples. First, a set of geometric indicators was designed to train a model and realize initial matching recognition. Then, a contextual metric was calculated based on the rough matching results, and geometric and contextual metrics were combined to train the model and realize relaxed matching recognition. Relaxed matching is an optimization process implemented after initial matching to recognize more relaxed matching relationships. In relaxed matching, a convex hull is used to recognize matching relationships besides 1:1, such as 1:n, m:1 and m:n. The experimental results showed that the proposed method outperformed indicator-weighted (weighted average) and learning-based matching methods, such as traditional SVMs and decision trees (DTs). The precision scores of the proposed model were 97.1%, 95% and 97.2% for the Wuhan (China), Beijing (China) and Richmond Hill (Canada) datasets, respectively. Furthermore, the proposed model identified the matching relationships of buildings with complex geometric features and high-density spatial distributions.

Symbolic Regression Enhanced Decision Trees for Classification Tasks

We introduce a conceptually simple yet effective method to create small, compact decision trees - by using splits found via Symbolic Regression (SR). Traditional decision tree (DT) algorithms partition a dataset on axis-parallel splits. When the true boundaries are not along the feature axes, DT is likely to have a complicated structure and a dense decision boundary. In this paper, we introduce SR-Enhanced DT (SREDT) - a method which utilizes SR to increase the richness of the class of possible DT splits. We evaluate SREDT on both synthetic and real-world datasets. Despite its simplicity, our method produces surprisingly small trees that outperform both DT and oblique DT (ODT) on supervised classification tasks in terms of accuracy and F-score. We show empirically that SREDTs decrease inference time (compared to DT and ODT) and argue that they allow us to obtain more explainable descriptions of the decision process. SREDT also performs competitively against state-of-the-art tabular classification methods, including tree ensembles and deep models. Finally, we introduce a local search mechanism to improve SREDT and evaluate it on 56 PMLB datasets. This mechanism shows improved performance on 77.2% of the datasets, outperforming DT and ODT. In terms of F-Score, local SREDT outperforms DT and ODT in 82.5% and 73.7% of the datasets respectively and in terms of inference time, local SREDT requires 25.8% and 26.6% less inference time than DT and ODT respectively.

Coevolutionary Algorithm for Building Robust Decision Trees under Minimax Regret

In recent years, there has been growing interest in developing robust machine learning (ML) models that can withstand adversarial attacks, including one of the most widely adopted, efficient, and interpretable ML algorithms—decision trees (DTs). This paper proposes a novel coevolutionary algorithm (CoEvoRDT) designed to create robust DTs capable of handling noisy high-dimensional data in adversarial contexts. Motivated by the limitations of traditional DT algorithms, we leverage adaptive coevolution to allow DTs to evolve and learn from interactions with perturbed input data. CoEvoRDT alternately evolves competing populations of DTs and perturbed features, enabling construction of DTs with desired properties. CoEvoRDT is easily adaptable to various target metrics, allowing the use of tailored robustness criteria such as minimax regret. Furthermore, CoEvoRDT has potential to improve the results of other state-of-the-art methods by incorporating their outcomes (DTs they produce) into the initial population and optimize them in the process of coevolution. Inspired by the game theory, CoEvoRDT utilizes mixed Nash equilibrium to enhance convergence. The method is tested on 20 popular datasets and shows superior performance compared to 4 state-of-the-art algorithms. It outperformed all competing methods on 13 datasets with adversarial accuracy metrics, and on all 20 considered datasets with minimax regret. Strong experimental results and flexibility in choosing the error measure make CoEvoRDT a promising approach for constructing robust DTs in real-world applications.

Novel Instance-Based Transfer Learning for Asphalt Pavement Performance Prediction

The deep learning method has been widely used in the engineering field. The availability of the training dataset is one of the most important limitations of the deep learning method. Accurate prediction of pavement performance plays a vital role in road preventive maintenance (PM) and decision-making. Pavement performance prediction based on deep learning has been widely used around the world for its accuracy, robustness, and automation. However, most of the countries in the world have not built their pavement performance historical database, which prevents preventive maintenance using the deep learning method. This study presents an innovative particle swarm optimization (PSO) algorithm-enhanced two-stage TrAdaBoost.R2 transfer learning algorithm, which could significantly increase the pavement performance prediction database. The Long-Term Pavement Performance (LTPP) database is used as the source domain data, and one of the highways in China is chosen as the target domain to predict pavement performance. The results show that the proposed PSO-Two-stage TrAdaBoost.R2 model has the highest accuracy compared with AdaBoost.R2 model and traditional regression decision tree model. The validation case study shows significant consistency between the predicted International Roughness Index (IRI) and the whole-year measurement data with an R2 of 0.7. This study demonstrates the great potential of the innovative instance-based transfer learning method in pavement performance prediction of a region’s lack of data. This study also contributes to other engineering fields that could greatly increase the universality of deep learning.

Fault Detection and Failure Rate Analysis of New Energy Vehicles Based on Decision Tree Algorithm

Abstract New energy vehicles are vital in promoting environmental protection and technological innovation. Fault detection still faces challenges during its operation, and efficient and accurate methods for fault diagnosis are urgently needed. This paper proposes a fault detection and analysis model based on a decision tree algorithm for the fault detection problem of new energy vehicles. The dataset applicable to the model is prepared by preprocessing in-vehicle network data, including data cleaning, integration, and other steps. Fault prediction can be realized after using C4.5 algorithms to construct a decision tree. With a precision of 82.26% on the test set, this model is highly accurate in fault detection, which is 1.23 percentage points higher than the traditional decision tree algorithm. The model’s effectiveness and efficiency in handling large-scale data were demonstrated by its training and testing on training sets of different sizes. Using the traditional algorithm, a training set of 80,000 data was used to reduce the model’s running time from 274,432 seconds to 249,269 seconds. This study provides a practical methodology for fault diagnosis of new energy vehicles, improving fault detection accuracy while optimizing computational efficiency. Real-time monitoring and timely maintenance of new energy vehicles require this.

Li‐Ion Battery State of Health Estimation Using Hybrid Decision Tree Model Optimized by Bayesian Optimization

Accurately battery state of health (SOH) estimation in electric vehicles (EVs) is crucial for optimal performance and safety. This article presents an in‐depth investigation into the utilization of an optimized decision tree (DT) model for precise SOH estimation. Two aging features (AFs) are extracted through incremental capacity analysis from the partial charging process, and their efficacy is thoroughly validated on both battery cells and packs. To capture the relationship between AFs and capacity degradation, a robust DT model is established. Furthermore, Bayesian optimization is employed to optimize the model hyperparameters, enhancing the learning and generalization capabilities of the DT model. To validate the optimized DT model, a battery management system (BMS) is developed using a hybrid programming approach with LabVIEW and MATLAB, enabling online SOH estimation. The BMS undergoes rigorous evaluation using publicly available aging datasets of battery cells, as well as private datasets of battery packs. The results demonstrate that the optimized DT model outperforms traditional DT models, back propagation neural networks, and support vector machine approaches in accurately estimating battery SOH across diverse datasets. This research contributes to the advancement of BMS and holds profound implications for the efficient utilization of batteries in EVs.

Algorithm Research for Bubble Detection in Capacitive Liquid Level Sensor of Fully Automated Immunoassay Analyzer

Fully automated immunoassay analyzers are integral to clinical testing, with liquid-level detection crucial to mitigate false measurements and cross-contamination. This study introduces a bubble detection algorithm based on the Random Forest machine learning model to address bubble detection in analyzers using capacitive sensors. The algorithm processes collected capacitance data, accurately identifying bubbles and ensuring sampling accuracy. Data preprocessing, feature extraction, and classification enable precise bubble detection. Compared to the traditional CART decision tree algorithm, this algorithm demonstrates superior bubble detection performance, indicating its significant research value and potential for liquid-level detection in fully automated immunoassay analyzers.

A secure energy trading in a smart community by integrating Blockchain and machine learning approach

ABSTRACT In today’s smart communities, small-scale energy systems are essential for sustainable development and efficient resource management. However, ensuring the confidentiality, safety, and accurate prediction of energy consumption patterns in energy trading is a major challenge. To address these issues, an innovative solution that synergistically combines two cutting-edge technologies: blockchain and machine learning is proposed. This paper unveils a novel approach that harmoniously merges blockchain with the Recalling-Enhanced Recurrent Neural Network (RERNN) to revolutionize energy trading systems called ‘Blockchain-Enhanced Energy Trading with Recalling-Enhanced Recurrent Neural Network (BET-RERNN).’ Data from IoT-enabled smart devices is securely stored in blockchain blocks, ensuring data integrity and immutability. Blockchain’s decentralized nature creates a trust-less environment for energy trading, protecting the privacy and anonymity of participants while maintaining transparency. At the heart of our system lies the advanced machine-learning capabilities of the RERNN model. By processing the data stored on the blockchain, RERNN accurately predicts optimal power generation for small-scale energy systems, enabling smart communities to make informed decisions and optimize their energy consumption. The BET-RERNN scheme provides a plethora of strengths. First, participants can securely engage in energy trading without compromising sensitive information, fostering a more resilient and efficient market. Second, blockchain technology ensures that all energy-related data is protected from tampering and unauthorized access, ensuring system reliability and trust. An in-depth comparison of RERNN’s performance to traditional General Regression Neural Network (GRNN) and Gradient Boost Decision Tree (GBDT) methods is conducted. To verify the strategy’s effectiveness, MATLAB simulations are employed, demonstrating its real-world applicability and scalability. By combining blockchain and machine learning, a secure and privacy-preserving smart community is established, promoting sustainable energy practices for a greener future.

HIE-EDT: Hierarchical interval estimation-based evidential decision tree

Decision tree algorithm, because of its strong interpretability and high algorithm efficiency, is widely used in the field of pattern recognition and classification. When the number of data samples is small and there is uncertainty in the data, it is difficult for the traditional decision tree algorithm to fully mine the effective information in the data. In this paper, we use the Dempster–Shafer framework to model data uncertainty and propose a hierarchical interval estimation method to improve decision tree algorithms. The proposed method constructs intervals through two methods of attribute boundary and mean square error estimation, which not only utilizes the characteristics of intervals to model the inaccuracy of data, but also constrains intervals from two aspects, narrowing the representation range of available information. By comparing with the classic decision tree algorithm and the decision tree algorithm based on single interval estimation, the proposed method can perform classification tasks robustly and accurately in different types of data under seven data sets.

Research on a Carbon Emission Prediction Method for Oil Field Transfer Stations Based on an Improved Genetic Algorithm—The Decision Tree Algorithm

The background of “dual carbon” is accelerating low-carbon transformation in the energy field, and oil field enterprises are facing challenges in energy conservation and emissions reduction for sustainable development. However, oil field gathering and transfer station systems, which are crucial components of the onshore transportation system, face challenges in energy conservation and emissions reduction. Therefore, it is necessary to predict the carbon emissions of oil field gathering and transfer station systems. To improve the accuracy of carbon emission prediction for the system, this study proposes an improved GA-decision tree (IGA-decision tree) algorithm. First, chaotic mapping was introduced to initialize the population, ensuring a uniform distribution of initial particles in the search space and enhancing population diversity. Second, the firefly perturbation strategy was employed to avoid the problem of genetic algorithms becoming trapped in local optima during the later stages of the search. The results show that the enhanced GA-decision tree algorithm effectively avoided being stuck in local optima while performing global searches. When predicting the carbon emissions of oil field gathering and transfer stations, the improved GA-decision tree (IGA-decision tree) algorithm outperformed traditional decision tree and GA-decision tree algorithms in terms of error and convergence efficiency. It achieved a root mean square error (RMSE) value of 74.5181 and a correlation coefficient (R2) of 0.99, indicating a high level of fitness and good convergence, as well as high prediction accuracy. This algorithm contributes to carbon accounting and energy conservation efforts in oilfield gathering and the transfer station system, filling the research gap in carbon emissions prediction for the system within the framework of energy internet projects.

Semantic decision Trees: A new learning system for the ID3-Based algorithm using a knowledge base

The decision tree (DT) is a well-established tool used to classify information. Its efficiency in accessing data and the accuracy of its outcomes is largely dependent on the characteristics of the input data and the efficacy of the construction processes and algorithms. In this study, we are introducing an innovative approach called the Semantic Decision Tree (SDT). The SDT builds upon the concept of the Iterative Dichotomiser 3 (ID3) algorithm. Our proposed SDT's first application involves utilizing a knowledge base, intentionally designed to support the DT construction process. The primary goal here is to effectively address the multi-value bias selection issue, a recurring problem found in a number of DT algorithms. Building on this, we adopt an ontology summarization technique to fine-tune the splitting condition of the DT. This adjustment is influenced by the attribute's importance value, extracted directly from the ontology. This process leads to the creation of nodes that align more precisely with the dataset than those generated by traditional DT methods. To validate the robustness and precision of our proposed SDT, we evaluated it across four distinct datasets. In addition, we benchmarked the classification accuracy of our SDT against several leading baseline algorithms; the traditional ID3, C4.5, CART, and the Mutual Information Decision Tree (MIDT). The findings demonstrate that our proposed SDT outperforms these industry-standard algorithms in terms of accuracy. Further in-depth qualitative analysis showed that the modifications in information gain resulted in DT structures that align more naturally with human decision-making logic.

EDST: a decision stump based ensemble algorithm for synergistic drug combination prediction

IntroductionThere are countless possibilities for drug combinations, which makes it expensive and time-consuming to rely solely on clinical trials to determine the effects of each possible drug combination. In order to screen out the most effective drug combinations more quickly, scholars began to apply machine learning to drug combination prediction. However, most of them are of low interpretability. Consequently, even though they can sometimes produce high prediction accuracy, experts in the medical and biological fields can still not fully rely on their judgments because of the lack of knowledge about the decision-making process.Related workDecision trees and their ensemble algorithms are considered to be suitable methods for pharmaceutical applications due to their excellent performance and good interpretability. We review existing decision trees or decision tree ensemble algorithms in the medical field and point out their shortcomings.MethodThis study proposes a decision stump (DS)-based solution to extract interpretable knowledge from data sets. In this method, a set of DSs is first generated to selectively form a decision tree (DST). Different from the traditional decision tree, our algorithm not only enables a partial exchange of information between base classifiers by introducing a stump exchange method but also uses a modified Gini index to evaluate stump performance so that the generation of each node is evaluated by a global view to maintain high generalization ability. Furthermore, these trees are combined to construct an ensemble of DST (EDST).ExperimentThe two-drug combination data sets are collected from two cell lines with three classes (additive, antagonistic and synergistic effects) to test our method. Experimental results show that both our DST and EDST perform better than other methods. Besides, the rules generated by our methods are more compact and more accurate than other rule-based algorithms. Finally, we also analyze the extracted knowledge by the model in the field of bioinformatics.ConclusionThe novel decision tree ensemble model can effectively predict the effect of drug combination datasets and easily obtain the decision-making process.

Adaptive neural decision tree for EEG based emotion recognition

An adaptive neural decision tree is investigated to recognize electroencephalogram (EEG) emotion signal with ability of intelligently selecting network structure. Firstly, to overcome lack of position information of the vectorized input EEG signal, the original one-dimensional EEG vector signal is cast into a two-dimensional matrix signal added with channel position information. Secondly, the depthwise convolution is adopted in transformers to deal with each channel by one convolution kernel, thus decreases the number of parameters. Then, to overcome model interpretability problem, an adaptive neural decision tree (ANT) based emotion recognition method is explored by embedding neural networks into the decision tree to perform feature extracting, path selecting, and label classifying. Further, ANT can automatically search optimized parameters by using the adaptive moment estimation algorithm, and explore tree architectures by using the exploration–exploitation trade-off reinforcement learning method to get the global optimal network structure without manually setting complex hyperparameters. Finally, by 5-fold cross-validation, binary, four-class and eight-class classification experiments are carried out on DEAP datasets, the average accuracy of the ANT algorithm are 99.14 ± 0.456%, 98.95 ± 0.84%, 97.58 ± 2.311%, respectively, which verifies the effectiveness of the proposed method, compared with the traditional decision tree method.

Real-time load forecasting model for the smart grid using bayesian optimized CNN-BiLSTM

A smart grid is a new type of power system based on modern information technology, which utilises advanced communication, computing and control technologies and employs advanced sensors, measurement, communication and control devices that can monitor the status and operation of various devices in the power system in real-time and optimise the dispatch of the power system through intelligent algorithms to achieve efficient operation of the power system. However, due to its complexity and uncertainty, how to effectively perform real-time prediction is an important challenge. This paper proposes a smart grid real-time prediction model based on the attention mechanism of convolutional neural network (CNN) combined with bi-directional long and short-term memory BiLSTM.The model has stronger spatiotemporal feature extraction capability, more accurate prediction capability and better adaptability than ARMA and decision trees. The traditional prediction models ARMA and decision tree can often only use simple statistical methods for prediction, which cannot meet the requirements of high accuracy and efficiency of real-time load prediction, so the CNN-BiLSTM model based on Bayesian optimisation has the following advantages and is more suitable for smart grid real-time load prediction compared with ARMA and decision tree. CNN is a hierarchical neural network structure containing several layers such as a convolutional layer, pooling layer and fully connected layer. The convolutional layer is mainly used for extracting features from data such as images, the pooling layer is used for the dimensionality reduction of features, and the fully connected layer is used for classification and recognition. The core of CNN is the convolutional operation, a locally weighted summation operation on the input data that can effectively extract features from the data. In the convolution operation, different features can be extracted by setting different convolution kernels to achieve feature extraction and classification of data. BiLSTM can capture semantic dependencies in both directions. The BiLSTM structure consists of two LSTM layers that process the input sequence in the forward and backward directions to combine the information in both directions to obtain more comprehensive contextual information. BiLSTM can access both the front and back inputs at each time step to obtain more accurate prediction results. It effectively prevents gradient explosion and gradient disappearance while better capturing longer-distance dependencies. The CNN-BiLSTM extracts features of the data and then optimises them by Bayes. By collecting real-time data from the power system, including power, load, weather and other factors, our model uses the features of CNN-BiLSTM to deeply learn real-time load data from smart grids and extract key features to achieve future load prediction. Meanwhile, the Bayesian optimisation algorithm based on the model can optimise the model’s hyperparameters, thus improving the model’s prediction performance. The model can achieve accurate prediction of a real-time power system load, provide an important reference for the dispatch and operation of the power system, and help optimise the operation efficiency and energy utilisation efficiency of the power system.