Search Results (4)

Customer retention is a key priority for mobile telecommunications companies, as acquiring new customers is significantly more costly than retaining existing ones. A major challenge in this field is predicting customer churn—users discontinuing services. Traditional predictive models such as rule-based systems often struggle with the complex, non-linear nature of customer behavior. To address this, we propose the use of deep learning techniques, specifically multi-layer perceptron (MLP) and radial basis function (RBF) networks, to improve the accuracy of churn predictions. However, while neural networks excel in predictive performance, they are often criticized for being “black-box” models, lacking interpretability. A real-world data set is considered, which originally contained information about 15,000 randomly selected clients. Various network structures and configurations are analyzed. The obtained results are compared with results generated using fuzzy rule-based and rough-set rule-based systems. The MLP model achieved an almost perfect accuracy of $0.999$ with an F-measure of $0.989$, outperforming traditional methods such as fuzzy rule-based and rough-set systems. Although the RBF model slightly lagged in accuracy, it demonstrated a superior recall of $0.993$, indicating better identification of potential churners. These results demonstrate that neural network models significantly enhance predictive performance in churn modeling. The interpretability of the model is also discussed since it bears significance in real applications. Our contribution lies in showing that deep learning methods significantly enhance churn prediction accuracy, though the challenge of model interpretability remains a critical area for future work. Full article

This paper presents (a) a novel hybrid learning method to train interval type-1 non-singleton type-3 fuzzy logic systems (IT3 NSFLS-1), (b) a novel method, named enhanced Wagner–Hagras (EWH) applied to IT3 NSFLS-1 fuzzy systems, which includes the level alpha 0 output to calculate the output $y$ alpha using the average of the outputs $y$ alpha $k$ instead of their weighted average, and (c) the novel application of the proposed methodology to solve the problem of transfer bar surface temperature prediction in a hot strip mill. The development of the proposed methodology uses the orthogonal least square (OLS) method to train the consequent parameters and the backpropagation (BP) method to train the antecedent parameters. This methodology dynamically changes the parameters of only the level alpha 0, minimizing some criterion functions as new information becomes available to each level alpha $k$. The precursor sets are type-2 fuzzy sets, the consequent sets are fuzzy centroids, the inputs are type-1 non-singleton fuzzy numbers with uncertain standard deviations, and the secondary membership functions are modeled as two Gaussians with uncertain standard deviation and the same mean. Based on the firing set of the level alpha 0, the proposed methodology calculates each firing set of each level alpha $k$ to dynamically construct and update the proposed EWH IT3 NSFLS-1 (OLS–BP) system. The proposed enhanced fuzzy system and the proposed hybrid learning algorithm were applied in a hot strip mill facility to predict the transfer bar surface temperature at the finishing mill entry zone using, as inputs, (1) the surface temperature measured by the pyrometer located at the roughing mill exit and (2) the time taken to translate the transfer bar from the exit of the roughing mill to the entry of the descale breaker of the finishing mill. Several fuzzy tools were used to make the benchmarking compositions: type-1 singleton fuzzy logic systems (T1 SFLS), type-1 adaptive network fuzzy inference systems (T1 ANFIS), type-1 radial basis function neural networks (T1 RBFNN), interval singleton type-2 fuzzy logic systems (IT2 SFLS), interval type-1 non-singleton type-2 fuzzy logic systems (IT2 NSFLS-1), type-2 ANFIS (IT2 ANFIS), IT2 RBFNN, general singleton type-2 fuzzy logic systems (GT2 SFLS), general type-1 non-singleton type-2 fuzzy logic systems (GT2 NSFLS-1), interval singleton type-3 fuzzy logic systems (IT3 SFLS), and interval type-1 non-singleton type-3 fuzzy systems (IT3 NSFLS-1). The experiments show that the proposed EWH IT3 NSFLS-1 (OLS–BP) system presented superior capability to learn the knowledge and to predict the surface temperature with the lower prediction error. Full article

Wood-based composites are increasingly used in the industry not only because of the shortage of solid wood, but above all because of the better properties, such as high strength and aesthetic appearance compared to wood. Medium-density fiberboard (MDF) is a wood-based composite that is widely used in the furniture industry. In this work, an attempt was made to predict the surface roughness of the machined MDF in the milling process based on acceleration signals from an industrial piezoelectric sensor installed in the cutting zone. The surface roughness parameter Sq was adopted for the evaluation and measurement of surface roughness. The surface roughness prediction was performed using a radial basis function (RBF) artificial neural network (ANN) and a Takagi–Sugeno––Kang (TSK) fuzzy model with subtractive clustering. In the research, as inputs to the ANNs and fuzzy model, the kinematic parameters of the cutting process and selected measures of the acceleration signal were adopted. At the output, the values of the surface roughness parameter Sq were obtained. The results of the experiments show that the surface roughness is influenced not only by the kinematic parameters of the cutting, but also by the vibrations generated during the milling process. Therefore, by combining information on the cutting kinematics parameters and vibration, the accuracy of the surface roughness prediction in the milling process of MDF can be improved. The use of TSK fuzzy modelling based on the subtractive clustering method for integrating the information from many acceleration signal measurements in the examined range of cutting conditions meant the surface roughness was predicted with high accuracy and high reliability. With the help of two tested artificial intelligence tools, it is possible to estimate the surface roughness of the workpiece with only a small error. When using a radial neural network, the root mean square error for estimating the value of the Sq parameter was 0.379 μm, while the estimation error based on fuzzy logic was 0.198 μm. The surface of the sample made with the cutting parameters v_c = 76 m/min and v_f = 1200 mm/min is characterized by a less concentrated distribution of ordinate densities, compared to the surface of the sample cut with lower feed rates but at the same cutting speed. The most concentrated distribution of ordinate density (for the cutting speed v_c = 76 m/min) is characterized by the surface, where the feed rate value was v_f = 200 mm/min, with 90% of the material concentrated in the profile height of 28.2 μm. When using an RBF neural network, the RMSE of estimating the value of the Sq parameter was 0.379 μm, while the estimation error based on fuzzy logic was 0.198 μm. Full article

The intelligent adjustment method of the shearer drum is the key technology to improve the intelligent level and safety degree of fully mechanized mining face. This paper proposes a shearer drum intelligent height adjustment model based on rough set significance attribute reduction (AR) and fuzzy rough radial basis function neural network (FRRBFNN) optimized by adaptive immune genetic algorithm (AIGA). The model first selects the parameters of shearer process monitoring based on the importance attribute reduction algorithm of rough set, and establishes the attribute reduction set of shearer operation characteristic parameters and the drum height decision rule base. Next, a fuzzy rough radial basis function neural network determined by the decision rule space is proposed. By introducing the fuzzy rough membership function as the connection weight, the network can accurately describe the complex nonlinear relationship between the working characteristic parameters of the attribute shearer and the drum height, and measure the uncertainty of the coal seam distribution. Finally, to further optimize the performance of FRRBFNN, the adaptive immune genetic algorithm is introduced to optimize its parameters, to build a high-precision shearer drum height prediction system. For the evaluation method of the model, we use three indicators: mean absolute error, mean absolute percentage error, and root mean square error. Based on the measured data in Yujialiang area, Shaanxi Province, the experimental results show that—compared with the FRRBFNN and support vector regression (SVR) models, a gated current neural network (GRU), a radial basis function neural network (RBF), the memory strengthen long short-term memory (MSLSTM) model, and the adaptive fuzzy reasoning Petri net (AFRPN)—the MAE of the AR-AIGA-FRBFNN model for predicting the height of the left and right rollers are 18.3 mm and 17.2 mm, respectively; the MAPE is 0.96% and 0.93%, respectively; and the RMSE is 21.2 mm and 22.4 mm, respectively. The AR-AIGA-FRBFNN model is therefore more effective than the other considered methods. Full article