Multi-scale and multi-indicator financial time series forecasting is a valuable requirement of quantitative analysis and intelligent decision-making. Nevertheless, the current research concentrates on single-step or univariate prediction, which is unable to completely meet the real-life needs of multi-step and multivariate prediction problems in complex financial time-series situations. Convolutional Long Short-Term Memory (ConvLSTM) networks combine convolutional feature extraction with temporal dependency modeling. This combination supports multivariate time series forecasting. Nevertheless, some important hyperparameters like network structure and learning rate are often tuned manually and therefore, it has low parameter configuration efficiency and high reliance on human knowledge. To overcome these challenges, this paper suggests an intelligent financial time series forecasting system that relies on ConvLSTM optimized by Island Algorithm (IA). The Island Algorithm is employed to optimize the important hyperparameters of the ConvLSTM model, including network structure, learning rate, training epochs, and convolution kernel size. By automatically searching for suitable parameter configurations, the proposed framework reduces the dependence on manual parameter tuning and improves the adaptability of ConvLSTM for financial time series forecasting tasks. In addition, three dimensions of evaluation are considered from three perspectives: forecasting performance, reliability of models, and value of practical application, which offer systematic dimensions for evaluating financial time series forecasting frameworks. The proposed method gives a systematic approach to utilize intelligent optimization algorithms to aid deep learning models in financial time series forecasting and provides references for intelligent forecasting applications in various financial scenarios.

Geometry-reconfigurable arrays offer additional spatial flexibility for cell-free integrated sensing and communication (ISAC), but unknown complex bistatic gains can limit their localization benefit. We derive a nuisance-aware equivalent Fisher information matrix that isolates identifiable angular information after gain elimination. The analysis shows that rank-one probing provides no transmit-angle information, even when a known-gain model predicts a positive value; rank at least two is necessary but not sufficient. Mapping angular information to position separates local aperture sensitivity from network angular diversity. We then jointly optimize communication and probing covariances and transmit/receive array shapes using semidefinite covariance optimization and safeguarded shape search, with deterministic rank-one recovery of communication covariances. A hybrid information model incorporates residual orientation and actuator errors to characterize calibration limits. Under a physical 28-GHz model, joint reconfiguration achieves a 5.51-dB paired Cramer-Rao bound gain over collapsed-geometry covariance-only design across 48 trials. Maximum-likelihood localization remains within 0.45 dB of the root-mean-square position error bound at the tested efficiency points. Orientation uncertainty substantially degrades localization, while covariance robustification offers little improvement in the evaluated setting. These results identify when geometry reconfiguration yields localization gains and how residual calibration errors limit their realization.

Accurate evaluation of radio access performance is essential for the design and optimization of ultra-dense fifth-generation (5G) networks, where the interactions between multiple users, base stations, and interference sources become highly complex. This paper proposes a comprehensive System-Level Simulation (SLS) framework for assessing the performance of multi-user multiple-input multiple-output (MIMO) systems in indoor hotspot environments, following the standardized 3rd generation partnership project (3GPP) InH_B channel model. The proposed framework integrates realistic network-level processes, including user equipment deployment, three-dimensional channel coefficient  generation, analog and digital beamforming, link adaptation, and block error rate prediction. Different from conventional SLS studies that often decouple receiver processing, channel state information (CSI) uncertainty, and link abstraction, the proposed framework provides a unified receiver-aware evaluation chain calibrated against 3GPP Phase 2 results. Both Maximal Ratio Combining (MRC) and Interference Rejection Combining (IRC) receivers are implemented to investigate the impact of imperfect CSI on system performance. Extensive simulations demonstrate that the proposed framework provides close agreement with the 3GPP Phase 2 calibration results, while enabling detailed receiver-level comparisons under practical CSI conditions. When CSI errors are considered, the IRC receiver achieves approximately 42% higher average cell throughput and more than 160% improvement in 5% user throughput compared to MRC, confirming its superior robustness to interference and channel estimation errors. The developed framework serves as a practical reference platform for future research on intelligent beamforming, adaptive MIMO receiver design, and machine learning-based link adaptation for beyond-5G and sixth-generation (6G) systems.