Sequence Research Articles

A Data-Driven Deep Learning Framework for Prediction of Traffic Crashes at Road Intersections

Traffic crash prediction (TCP) is a fundamental problem for intelligent transportation systems in smart cities. Improving the accuracy of traffic crash prediction is important for road safety and effective traffic management. Owing to recent advances in artificial neural networks, several new deep-learning models have been proposed for TCP. However, these works mainly focus on accidents in regions, which are typically pre-determined using a grid map. We argue that TCP for roads, especially for crashes at or near road intersections which account for more than 50% of the fatal or injury crashes based on the Federal Highway Administration, has a significant practical and research value and thus deserves more research. In this paper, we formulate TCP at Road Intersections as a classification problem and propose a three-phase data-driven deep learning model, called Road Intersection Traffic Crash Prediction (RoadInTCP), to predict traffic crashes at intersections by exploiting publicly available heterogeneous big data. In Phase I we extract discriminative latent features called topological-relational features (tr-features), of intersections using a neural network model by exploiting topological information of the road network and various relationships amongst nearby intersections. In Phase II, in addition to tr-features which capture some inherent properties of the road network, we also explore additional thematic information in terms of environmental, traffic, weather, risk, and calendar features associated with intersections. In order to incorporate the potential correlation in nearby intersections, we utilize a Graph Convolution Network (GCN) to aggregate features from neighboring intersections based on a message-passing paradigm for TCP. While Phase II serves well as a TCP model, we further explore the signals embedded in the sequential feature changes over time for TCP in Phase III, by exploring RNN or 1DCNN which have known success on sequential data. Additionally, to address the serious issues of imbalanced classes in TCP and large-scale heterogeneous big data, we propose an effective data sampling approach in data preparation to facilitate model training. We evaluate the proposed RoadInTCP model via extensive experiments on a real-world New York City traffic dataset. The experimental results show that the proposed RoadInTCP robustly outperforms existing methods.

Understanding price momentum, market fluctuations, and crashes: insights from the extended Samuelson model

Although momentum strategies result in abnormal profitability, thereby challenging the efficient market hypothesis (EMH), concerns persist regarding their reliability due to their significant volatility and susceptibility to substantial losses. In this study, we investigate the limitations of these strategies and propose a solution. Our literature review reveals that the volatile profits are due to statistical analyses that assume the persistence of past patterns, leading to unreliable results in out-of-sample scenarios when underlying mechanisms evolve. Statistical analysis, the predominant method in financial economics, often proves inadequate in explaining market fluctuations and predicting crashes. To overcome these limitations, a paradigm shift towards dynamic approaches is essential. Drawing inspiration from three groundbreaking economists, we introduce the extended Samuelson model (ESM), a dynamic model that connects price changes to market participant actions. This paradigm transition uncovers several significant findings. First, timely signals indicate momentum initiations, cessations, and reversals, validated using S&P 500 data from 1999 to 2023. Second, ESM predicts the 1987 Black Monday crash weeks in advance, offering a new perspective on its underlying cause. Third, we classify sequential stock price data into eight distinct market states, including their thresholds for transitions, laying the groundwork for market trend predictions and risk assessments. Fourth, the ESM is shown to be a compelling alternative to EMH, offering potent explanatory and predictive power based on a single, realistic assumption. Our findings suggest that ESM has the potential to provide policymakers with proactive tools, enabling financial institutions to enhance their risk assessment and management strategies.

Regularities of the theory of quasi-geodesic mappings of special parabolic spaces

We study quasi-geodesic mappings (QGM) of generalized-recurrent-parabolic spaces f: (Vn, gij, Fih) → (V'n, g'ij, Fih). QGM can be of two types: general and canonical. This article examines the QGM of the general type. Earlier, we considered the fundamental questions of the theory of QGM of generalized-recurrent-parabolic spaces. We proved theorems that allow for any generalized-recurrent-parabolic space (Vn, gij, Fih) to either find all spaces (V'n, g'_{ij}, Fih) on which Vn admits QGM of the general form, or prove that there are no such spaces. In this article, we constructed a Γ-transformation that makes it possible to obtain from a pair of generalized-recurrent-parabolic spaces that are in a quasi-geodesic mapping, an infinite sequence of pairs of other generalized-recurrent-parabolic spaces, which are also in a quasi-geodesic mapping.

ENHANCING 〖PM〗_(2.5) PREDICTION IN KEMAYORAN DISTRICT, DKI JAKARTA USING DEEP BILSTM METHOD

Worldwide air pollution is a concern, and this is especially true in Indonesia, where most people breathe air that is more contaminated than recommended by the WHO. The concentration of presents notable health hazards. The respiratory system is the primary route of absorption for , allowing it to enter the lung alveoli and enter the bloodstream. Given the significant health risks associated with exposure, accurate forecasting methods are crucial to anticipate and mitigate its effects. Traditional forecasting methods like ARIMA have limitations in handling non-linear and complex patterns. Therefore, an accurate machine learning method is needed to improve forecasting performance. This research employs Deep Bidirectional Long-Short Term Memory (BiLSTM), a deep learning model particularly suited for time series forecasting due to its ability to capture both past and future dependencies in sequential data. To achieve accurate and precise forecasts for predicting concentration levels in Kemayoran District in November , 2023 (24 hours), this research utilized hourly concentration data from May until October , 2023, using Deep BiLSTM. The outcomes demonstrated the efficiency of the model, attaining a Mean Absolute Percentage Error (MAPE) of 17.1540% (training) and 14.2862% (testing) with an 80:20 data split. The optimal parameters, which comprised 24 timesteps, Adam optimizers with a learning rate of 0.001, 16 batch sizes, 1000 epochs, and ReLU activation functions across multiple BiLSTM layers, showcased the model’s effectiveness in forecasting the concentration in Kemayoran District, DKI Jakarta, on November , 2023.

Physics-Based Synthetic Data Model for Automated Segmentation in Catalysis Microscopy.

In catalysis research, the amount of microscopy data acquired when imaging dynamic processes is often too much for nonautomated quantitative analysis. Developing machine learned segmentation models is challenged by the requirement of high-quality annotated training data. We thus substitute expert-annotated data with a physics-based sequential synthetic data model. We study environmental scanning electron microscopy (ESEM) data collected from isopropanol oxidation to acetone over cobalt oxide as an example. Upon applying a temperature program during the reaction a phase transition occurs, reducing the catalyst selectivity toward acetone. This is accompanied on the micrometer ESEM scale by the formation of cracks between the pores of the catalyst surface. We aim to generate synthetic data to train a neural network capable of semantic segmentation (pixel-wise labeling) of this ESEM data. This analysis will lead to insights into this phase transition. To generate synthetic data that approximates this transition, our algorithm composes the ESEM images of the room-temperature catalyst with dynamically evolving synthetic cracks satisfying physical construction principles, gathered from qualitative knowledge accessible in the ESEM data. We mimic the surface crack growth propagation along surface paths, avoiding close vicinity to nearby pores. This physics-based approach results in a lowered rate of false positives compared to a random approach.

COMPARING FORECASTS OF AGRICULTURAL SECTOR EXPORT VALUES USING SARIMA AND LONG SHORT-TERM MEMORY MODELS

Indonesia's agricultural sector plays a crucial role in the national economy, with significant export potential and supporting the livelihoods of the majority of the population. As part of the government's vision to make Indonesia the world's food barn by 2045, increasing the volume and value of agricultural product exports is a primary focus, making export value forecasting essential for supporting strategic decision-making. Sequential data analysis is an important approach in analyzing data collected over a specific period. This study aims to compare two popular methods in forecasting the export value of the agricultural sector, namely the Seasonal AutoRegressive Integrated Moving Average (SARIMA) model and the Long Short-Term Memory (LSTM) model. Monthly agricultural export data from West Java Province from January 2013 to February 2024 were used as the dataset. The best SARIMA model generated was (1,1,1)(0,1,1)12, while the optimal parameters for the LSTM model were neuron: 50, dropout rate: 0.3, number of layers: 2, activation function: relu, epochs: 500, batch size: 64, optimizer: Adam, and learning rate: 0.01. Evaluation was performed using the Root Mean Squared Error (RMSE) method to measure the accuracy of both models in forecasting the export value of the agricultural sector. The results showed that the LSTM model outperformed the SARIMA model, with a lower RMSE value (SARIMA: 4182.133 and LSTM: 1939.02). This research provides valuable insights for decision-makers in planning agricultural sector export strategies in the future. With this comparison, it is expected to provide better guidance in selecting forecasting methods suitable for the characteristics of the data.

Long-term Prospective Comparative Analysis of Ototoxic and Survival Outcomes of Sequential Boost and Simultaneous Integrated Boost of Volumetric Modulated Arc Therapy for Head-Neck Carcinomas.

To compare the ototoxicity and survival in head and neck carcinoma patients treated with sequential (SEQ) and simultaneous integrated boost (SIB) of volumetric modulated arc therapy (VMAT). This long-term prospective study enrolled patients with histologically confirmed head and neck carcinoma, all receiving VMAT treatment. Audiological assessments were done using various tests at baseline, two weeks, treatment completion, six months, and 12 months. The changes in bone conduction pure tone thresholds were correlated with cochlear dose, comparing SEQ and SIB plans. We also investigated other significant late toxicities that led to dysphagia, voice changes, and xerostomia. Survival was assessed with the Kaplan-Meier analysis. The study included 93 patients (186 ears), 40 receiving radiation alone and 53 undergoing chemoradiation. Baseline hearing levels for the right and left ears were 13.3±2.3 dB and 14.2±1.5 dB. After 12 months of radiation, levels were 18.5±2.4 dB and 19.11±1.9 dB, respectively. No significant changes were observed between SEQ and SIB plans, but high-frequency shifts occurred. The cochlea tolerated up to 28 Gy without hearing loss in the radiation-alone group but showed loss at 9 Gy when combined with cisplatin chemotherapy. The maximum dose (Dmax) and the mean dose (Dmean) of pharyngeal constrictor muscles predicted dysphagia. No significant SEQ vs. SIB differences were found in late toxicity or survival outcomes. Modern radiotherapy techniques like VMAT adhere to cochlear dose limits. No significant differences were found between SEQ and SIB plans in sensorineural hearing loss, late toxicity, or survival, making both suitable for head and neck carcinoma treatment.

Deep Learning Algorithm-Based Appearance Detection of Cigarette Strip Packs

Each local feature in the appearance image of cigarette packs is a key element to reflect the corresponding brand information. If only a single convolutional neural network is used, the context information of sequence data may be lost, resulting in an insufficient grasp of the overall information. In order to realize the deep-level feature extraction of the appearance of cigarette packs and realize the appearance detection of cigarette packs with higher accuracy and speed, a new method for the appearance detection of cigarette packs was proposed by combining the convolutional neural network and the cyclic neural network methods in the deep learning algorithm. The image acquisition card is used to collect the appearance images of cigarette packs, and contrast enhancement and rotation correction are performed on the collected images to effectively improve their quality and provide a good guarantee for subsequent feature extraction and detection. The preprocessed cigarette package image is input into the convolutional neural network to realize the deep-level feature extraction of the cigarette package appearance image. The cigarette package appearance features’ output from the convolutional neural network is input into the short-term and long-term memory unit, as well as the gate-controlled cyclic unit, of the corresponding recurrent neural network, in order to process time sequence information based on the efficient extraction of details from the image, retain the sequence and context information of the input data, and ultimately achieve accurate detection of the cigarette package appearance. Through experimental analysis, this method can effectively identify the appearance defects of cigarette packs, mark them immediately, and present them in an intuitive way, so that staff can quickly locate the problem and take corresponding measures. The method can detect appearance defects larger than [Formula: see text] with high accuracy. For various appearance defects, the detection rate can be guaranteed to be over 98%, providing strong support for quality control and product upgrading in the tobacco industry.

Design of an Algorithm for Rapid Estimation of the Unit Cost of Concrete in Construction Projects

The estimation of the concrete unit cost in construction engineering plays a vital role in controlling construction costs and ensuring timely project completion. To achieve this, a fast estimation algorithm for concrete unit costs is proposed. To ensure accurate estimation and reduce the computational burden on the convolutional neural network (CNN) for rapid estimation, the factors influencing concrete costs are identified and arranged using the interpretive structural modeling (ISM) method. The factors affecting the concrete unit cost of high-rise construction are selected as the input data sequence for the CNN model. The feature map of the input data is extracted through the convolution layer. After applying the pooling operation to the feature map, the processed data is passed into the fully connected layer through the final pooling layer, where the final calculation is performed to obtain the estimated concrete unit cost. Experimental results indicate that to ensure fast convergence and minimize estimation errors in CNN estimation, optimal network parameters are determined through multiple experiments. The best configuration includes a [Formula: see text] convolution kernel and 11 convolutional cores. This algorithm, which uses the ISM method to select influential factors, achieves high accuracy in estimating the concrete unit cost. In actual engineering applications, the error between the estimated and actual results is minimal, with a maximum difference of only 1 yuan.

AI-based models to predict decompensation on traumatic brain injury patients.

Traumatic Brain Injury (TBI) is a form of brain injury caused by external forces, resulting in temporary or permanent impairment of brain function. Despite advancements in healthcare, TBI mortality rates can reach 30%-40% in severe cases. This study aims to assist clinical decision-making and enhance patient care for TBI-related complications by employing Artificial Intelligence (AI) methods and data-driven approaches to predict decompensation. This study uses learning models based on sequential data from Electronic Health Records (EHR). Decompensation prediction was performed based on 24-h in-mortality prediction at each hour of the patient's stay in the Intensive Care Unit (ICU). A cohort of 2261 TBI patients was selected from the MIMIC-III dataset based on age and ICD-9 disease codes. Logistic Regressor (LR), Long-short term memory (LSTM), and Transformers architectures were used. Two sets of features were also explored combined with missing data strategies by imputing the normal value, data imbalance techniques with class weights, and oversampling. The best performance results were obtained using LSTMs with the original features with no unbalancing techniques and with the added features and class weight technique, with AUROC scores of 0.918 and 0.929, respectively. For this study, using EHR time series data with LSTM proved viable in predicting patient decompensation, providing a helpful indicator of the need for clinical interventions.

Tendon Anomaly Identification in Prestressed Concrete Beams Based on an Advanced Monitoring MEMS and Data-Driven Detection of Structural Damage.

The growing importance of state assessments in civil engineering has led to intensive research into the development of damage identification methods based on vibrations. Natural frequencies and modal shapes have garnered great interest because modal parameters are invariant of structure. Moreover, thanks to the global nature of modal parameters, their variations are not limited to the location of the damage. This is an important advantage that offers the opportunity to identify damage with sensors whose position does not have to coincide with the damaged area. The integration of MEMS sensors into structural health monitoring (SHM) systems offers a promising approach to long-term structural maintenance, especially in large-scale infrastructure. This paper presents an anomaly detection technique that analyzes raw sequential data within a statistical framework to detect damage that causes prestress loss of the tendon by exploiting a distributed monitoring system composed of six high-performance MEMS sensors. The proposed system is preliminarily evaluated to identify the frequency of the first mode, and then the proposed methodology is validated on acceleration data collected on a 240 cm beam in three different damage configurations, achieving a high detection accuracy and showing that its output can also evaluate the damage localization.

Adaptive Oversampling via Density Estimation for Online Imbalanced Classification

Online learning is a framework for processing and learning from sequential data in real time, offering benefits such as promptness and low memory usage. However, it faces critical challenges, including concept drift, where data distributions evolve over time, and class imbalance, which significantly hinders the accurate classification of minority classes. Addressing these issues simultaneously remains a challenging research problem. This study introduces a novel algorithm that integrates adaptive weighted kernel density estimation (awKDE) and a conscious biasing mechanism to efficiently manage memory, while enhancing the classification performance. The proposed method dynamically detects the minority class and employs a biasing strategy to prioritize its representation during training. By generating synthetic minority samples using awKDE, the algorithm adaptively balances class distributions, ensuring robustness in evolving environments. Experimental evaluations across synthetic and real-world datasets demonstrated that the proposed method achieved up to a 13.3 times improvement in classification performance over established oversampling methods and up to a 1.66 times better performance over adaptive rebalancing approaches, while requiring significantly less memory. These results underscore the method’s scalability and practicality for real-time online learning applications.

Large sets containing no copies of a given infinite sequence

Large sets containing no copies of a given infinite sequence

An Integrated Framework for Infectious Disease Control Using Mathematical Modeling and Deep Learning.

Infectious diseases are a major global public health concern. Precise modeling and prediction methods are essential to develop effective strategies for disease control. However, data imbalance and the presence of noise and intensity inhomogeneity make disease detection more challenging. Goal: In this article, a novel infectious disease pattern prediction system is proposed by integrating deterministic and stochastic model benefits with the benefits of the deep learning model. Results: The combined benefits yield improvement in the performance of solution prediction. Moreover, the objective is also to investigate the influence of time delay on infection rates and rates associated with vaccination. Conclusions: In this proposed framework, at first, the global stability at disease free equilibrium is effectively analysed using Routh-Haurwitz criteria and Lyapunov method, and the endemic equilibrium is analysed using non-linear Volterra integral equations in the infectious disease model. Unlike the existing model, emphasis is given to suggesting a model that is capable of investigating stability while considering the effect of vaccination and migration rate. Next, the influence of vaccination on the rate of infection is effectively predicted using an efficient deep learning model by employing the long-term dependencies in sequential data. Thus making the prediction more accurate.

Combining sequential extractions with bulk and micro X-ray spectroscopy to elucidate iron and phosphorus speciation in sediments of an iron-treated peat lake.

In shallow lakes, mobilization of legacy phosphorus (P) from the sediments can be the main cause for persisting eutrophication after reduction of external P input. In-lake remediation measures can be applied to reduce internal P loading and to achieve ecosystem recovery. The eutrophic shallow peat lake Terra Nova (The Netherlands) was treated with iron (Fe) to enhance P retention in the sediment. This treatment, however, intensified seasonal internal P loading. An earlier study suggested that Fe addition led to increased P binding by easily-reducible Fe(III) associated with organic matter (OM), which readily releases P when bottom waters turn hypoxic. In this complementary study, bulk and micro Fe K-edge and P K-edge X-ray absorption spectroscopy and micro-focused X-ray fluorescence spectroscopy were applied to characterize the P hosting Fe(III) pool. Combined with sequential extraction data, the synchrotron X-ray analyses revealed that a continuum of co-precipitates of Fe(III) with calcium, phosphate, manganese and organic carbon within the OM matrix constitutes the reducible Fe(III) pool. The complementary analyses also shed new light on the interpretation of sequential extraction results, demonstrating that pyrite was not quantitatively extracted by nitric acid (HNO3) and that most of the Fe(II) extracted by hydrochloric acid (HCl) originated from phyllosilicate minerals. Formation of an amorphous inorganic-organic co-precipitate upon Fe addition constitutes an effective P sink in the studied peaty sediments. However, the high intrinsic reactivity of this nanoscale co-precipitate and its fine distribution in the OM matrix makes it very susceptible to reductive dissolution, leading to P remobilization under reducing conditions.

Dynamic FeOx/FeWOx nanocomposite memristor for neuromorphic and reservoir computing.

Memristors are crucial in computing due to their potential for miniaturization, energy efficiency, and rapid switching, making them particularly suited for advanced applications such as neuromorphic computing and in-memory operations. However, these tasks often require different operational modes-volatile or nonvolatile. This study introduces a forming-free Ag/FeOx/FeWOx/Pt nanocomposite memristor capable of both operational modes, achieved through compliance current (CC) adjustment and structural engineering. Volatile switching occurs at low CC levels (<500 μA), transitioning to nonvolatile at higher levels (mA). Operating at extremely low voltages (<0.2 V), this memristor exhibits excellent uniformity, data retention, and multilevel switching, making it highly suitable for high-density data storage. The memristor successfully mimics fundamental biological synapse functions, exhibiting potentiation, depression, and spike-rate dependent plasticity (SRDP). It effectively emulates transitions from short-term memory (STM) to long-term memory (LTM) by varying pulse characteristics. Leveraging its volatile switching and STM features, the memristor proves ideal for reservoir computing (RC), where it can emulate dynamic reservoirs for sequence data classification. A physical RC system, implemented using digits 0 to 9, achieved a recognition rate of 93.4% in off-chip training with a deep neural network (DNN), confirming the memristor's effectiveness. Overall, the dual-mode switching capability of the Ag/FeOx/FeWOx/Pt memristor enhances its potential for AI applications, particularly in temporal and sequential data processing.

Forecasting A Major Banking Corporation Stock Prices Using LSTM Neural Networks

The increasing complexity of stock market predictions necessitates advanced computational techniques to address the unique challenges posed by financial data's non-linear and volatile nature. This study aims to leverage Long Short-Term Memory (LSTM) neural networks to accurately forecast stock prices, using historical data collected from a major banking corporation as a primary source. The LSTM model excels at processing sequential time-series data, allowing it to predict monthly stock closing prices over a one-year horizon with a high degree of precision. Our findings indicate a Root Mean Squared Error (RMSE) of 3.2, underscoring the model's efficiency and reliability in financial forecasting tasks. The novelty of this research lies in the systematic incorporation of preprocessing techniques and fine-tuned hyperparameters to optimize model performance. Furthermore, this study explores the practical implications of implementing LSTM models in real-world trading scenarios, analyzing their adaptability to dynamic market conditions and their potential integration into automated trading systems. These findings contribute to the growing body of knowledge in financial analytics and demonstrate the viability of machine learning-based solutions for accurate and robust market predictions.

A CNN‐LSTM Method Based on Voltage Deviation for Predicting the State of Health of Lithium‐Ion Batteries

ABSTRACTEnsuring the accurate estimation of the state of health (SOH) of lithium‐ion batteries (LIBs) is essential for the reliability and safe operation of battery management systems. The prediction of SOH has witnessed significant advancements recently, largely propelled by the powerful nonlinear modeling capabilities of deep learning. Despite these advancements, the intricate nature of the battery degradation process poses a challenge in accurately simulating it using measurement data. In this paper, we introduce a novel approach by focusing on the charging voltage deviation, which is defined as the discrepancy between the charging voltage and its average value over each charge/discharge cycle. This deviation is rooted in the electrochemical reactions that lead to capacity decay and voltage fluctuations. We propose a convolutional neural network‐long short‐term memory (CNN‐LSTM) hybrid framework aimed at estimating the SOH of the battery. For each charge/discharge cycle, a conventional CNN is employed to extract key capacity features from sequential charging data, encompassing voltage deviation, current, and charging duration. Following this, an LSTM network is leveraged to build the long‐term dependencies of battery capacities, facilitating the SOH prediction process. The experimental results indicate that our model not only simplifies the computational complexity but also significantly enhances the precision of SOH predictions. This innovative approach holds promise for the advancement of battery management systems, ensuring their continued reliability and safety.

Advancements in mobile AI: A machine learning-driven approach to enhance user experience and functionality

Artificial intelligence (AI) and mobile technology have synergized to achieve remarkable advancements across various sectors, transforming user interactions and significantly boosting mobile device performance. This study examines how integrating machine learning (ML) techniques into mobile applications enhances user satisfaction, productivity, and security. By deploying predictive models and real-time analysis directly on mobile devices, our approach reduces latency and personalizes experiences to better adapt to user behavior. Key findings reveal that the Hybrid CNN-LSTM model achieves superior accuracy (93.8%), precision (92.1%), and F1-score (91.5%) compared to standalone CNN or LSTM models, with a manageable latency of 140 ms, making it optimal for tasks requiring both image and sequential data processing. Additionally, applying optimization techniques like knowledge distillation reduces model size by 40% and latency by 25%, enhancing device efficiency without compromising performance. This study confirms that mobile-based AI equipped with advanced, autonomous decision-making capabilities enhances user-centric services and application responsiveness. Through experimental evaluations, this paper underscores the transformative impact of ML on mobile technology and proposes strategies to further integrate AI into the mobile ecosystem.

Customer churn prediction model based on hybrid neural networks

In today’s competitive market environment, accurately identifying potential churn customers and taking effective retention measures are crucial for improving customer retention and ensuring the sustainable development of an organization. However, traditional machine learning algorithms and single deep learning models have limitations in extracting complex nonlinear and time-series features, resulting in unsatisfactory prediction results. To address this problem, this study proposes a hybrid neural network-based customer churn prediction model, CCP-Net. In the data preprocessing stage, the ADASYN sampling algorithm balances the sample sizes of churned and non-churned customers to eliminate the negative impact of sample imbalance on the model performance. In the feature extraction stage, CCP-Net uses Multi-Head Self-Attention to learn the global dependencies of the input sequences, combines with BiLSTM to capture the long-term dependencies in the sequential data, and uses CNN to extract the local features, and ultimately generates the prediction results. Experimental results of cross-validation on Telecom, Bank, Insurance, and News datasets show that CCP-Net outperforms the comparison algorithms in all performance metrics. For example, CCP-Net achieves a Precision of 92.19% on the Telecom dataset, 91.96% on the Bank dataset, 95.87% on the Insurance dataset, and 95.12% on the News dataset, which compares to other hybrid neural network models, the performance improvement of CCP-Net ranges from 1% to 3%. These results indicate that the design of the CCP-Net model effectively improves the accuracy and robustness of churn prediction, enabling it to be widely applied to different industries, especially in the financial, telecommunication, and media fields, to provide more comprehensive and effective churn management strategies for enterprises.