Vol. 11, No. 04 [April 2025]

By combining predictive analytics, machine learning, customer segmentation, and relationship management practices, this study looks at ways to increase customer retention. It divides approaches into four groups: relationship management, machine learning, customer segmentation, and predictive and prescriptive analytics. Superior prediction accuracy and profitability optimization were shown by sophisticated models such as Predict-and-Optimize (PnO) and machine learning algorithms like Random Forest and Light Gradient Boosting Machine (LGBM). Targeted campaigns for a variety of client profiles were made possible by segmentation frameworks like Time-Frequency-Monetary (TFM) and Recency-Frequency-Monetary (RFM), which improved personalization. Loyalty and satisfaction were also greatly enhanced by relationship marketing, trust-building programs, and better service quality. The study emphasizes how combining data-driven insights with customer-centric tactics may work in concert to lower attrition, boost engagement, and boost long-term profitability. These results provide useful information for companies, as well as policymakers, industry regulators, and educators designing curricula for marketing, data science, and customer relationship management programs.
Keywords: customer retention, predictive analytics, machine learning, customer segmentation, relationship management, churn prevention, personalization, data-driven strategies, customer satisfaction, loyalty programs.

This new study includes an analysis and forecast of the time series of daily ozone maxima for 2024 and daily concentrations, plus a trend from 2010 to 2024 for the case of daily ozone maxima in Mexico City, Machine Learning and Deep Learning algorithms are used to study the behavior, as well as the classic ARIMA method for evaluation of the forecast model. RMSE, MSE, MAE and MAPE were used.
Keywords: Machine Learning, Deep Learning, Arima, Time Series, Ozone, Bivariate Probability Distributions, Stochastic Gaussian Mixture, Neural Networks.

To address low efficiency in manual pea seeding and poor adaptability of large machinery in China, a small-scale intelligent precision pea seeder suitable for small farmlands was designed. Integrating automated control and Internet of Things (IoT) technology, this device enables integrated operations of precise seeding, watering, fertilizing, and soil covering. Through a three-rail seeding mechanism, a dial-type precision seed metering device, and remote control via a smart APP, it solves problems such as uneven seeding spacing, high seed damage rate, and insufficient intelligence in traditional seeders. Test results show that the seeder significantly improves the efficiency and quality of pea seeding, providing an effective solution for mechanized pea cultivation in hilly and small-scale farmlands.
Keywords: Pea seeder; Precision seeding; Intelligence; Small-scale agriculture; Integrated operations

Predicting concrete's thermal expansion coefficient (CTE) is important for improving its performance in different environmental conditions. Traditional methods often fail to capture the complex relationships between concrete composition and thermal properties. Machine learning offers a better way to solve this problem by using data-driven models. This study used 192 datasets that included information on various concrete mixtures and their properties. Both simple and advanced machine learning models were tested to predict CTE. Simple methods, like Multiple Linear Regression (MLR) and Support Vector Regression (SVR), provided basic predictions. Advanced methods, such as AdaBoost, XGBoost, Bagging, and Random Forest, were used to handle the complex and non-linear nature of the data. These models were evaluated using methods like cross-validation and feature importance analysis to check their accuracy and reliability. When comparing the results, the study found that simpler models like MLR could only give limited insights. In contrast, advanced models, especially Random Forest, performed much better. The Random Forest model was highly accurate, identifying key factors affecting CTE and correctly predicting patterns in the data. It also reduced the number of experiments needed for accurate predictions, saving time and resources. This research shows how machine learning can be used to better predict and understand concrete properties. It highlights Random Forest as a powerful tool for accurately modeling CTE and other mechanical properties of concrete while reducing costs and improving efficiency
Keywords: Coefficient of Thermal Expansion, Concrete Performance, Machine Learning Models, Multiple Linear Regression, Support Vector Regression

To address the challenges of high energy consumption and poor stability in traditional obstacle-crossing robots operating in complex terrains, this study proposes a novel two-wheel self-balancing robot integrated with a cam-spring composite energy-storage mechanism. The robot employs a torque servo motor to drive a cam for compressing springs, thereby storing elastic potential energy. Combined with a PID-based self-balancing control system, the robot achieves an efficient obstacle-crossing process characterized by "energy pre-storage, instantaneous release, and dynamic balance." The modular mechanical design features an optimized cam profile to convert rotational motion into nonlinear linear motion, while the chassis utilizes high-precision encoders and a nine-axis sensor for real-time posture adjustment. Experimental results demonstrate a 98% obstacle-crossing success rate, a 40% improvement in motion smoothness, and a 35% reduction in energy consumption compared to conventional solutions. This work provides a new pathway for designing lightweight, high-reliability obstacle-crossing robots. A functional prototype has been developed, and a utility model patent has been filed.
Keywords: Obstacle-crossing robot; Energy-storage mechanism; Cam-spring composite transmission; Self-balancing control; Anti-lock mechanism

This research designs and develops a new security model for cloud computing to enhance Information Security using Blowfish Encryption and Advance Encryption Standard Algorithms. The proposed security model will improves the polytechnic performance by using minimum resources and management support, with a shared network, valuable resources as it provides a mechanism through which communication can be protected as well as hides the confidential information from unauthorized users. In this model, a combination of Advanced Encryption Standard (AES), Blowfish algorithm and Short Message Service (SMS) is implemented. Their combined features provides three way security i.e. confidentiality, authentication and verification. The proposed security system addressed issues of privacy, confidentiality, security and integrity of data stored in the cloud. This research will be of benefit to the polytechnic management, staff and the students of Computer Science Department. The resulting application is designed using Object Oriented Analysis and Design Method (OOADM) and is implemented using C# programming language and MYSQL data base.
Keywords: Information Security, Advance Encryption, Cloud, Application