Deep reinforcement learning (DRL) has been applied in financial portfolio management to improve returns in changing market conditions. However, unlike most fields where DRL is widely used, the stock market is more volatile and dynamic as it is affected by several factors such as global events and investor sentiment. Therefore, it remains a challenge to construct a DRL-based portfolio management framework with strong return capability, stable training, and generalization ability. This study introduces a new framework utilizing the Memory Instance Gated Transformer (MIGT) for effective portfolio management. By incorporating a novel Gated Instance Attention module, which combines a transformer variant, instance normalization, and a Lite Gate Unit, our approach aims to maximize investment returns while ensuring the learning process's stability and reducing outlier impacts. Tested on the Dow Jones Industrial Average 30, our framework's performance is evaluated against fifteen other strategies using key financial metrics like the cumulative return and risk-return ratios (Sharpe, Sortino, and Omega ratios). The results highlight MIGT's advantage, showcasing at least a 9.75% improvement in cumulative returns and a minimum 2.36% increase in risk-return ratios over competing strategies, marking a significant advancement in DRL for portfolio management.

Many real-world applications require prediction that takes the most advantage of data. Classic data mining mechanisms tend to feed a prediction model pivotal data to achieve a promising result, which needs to be adjusted in different application scenarios. Recent studies have shown the potential of I Ching mechanism to improve prediction capacity. However, the I Ching prediction mechanism has several issues, including underutilized I Ching knowledge and incomplete data conversion. To address these issues, the authors designed a model to leverage I Ching knowledge and transform traditional I Ching prediction processing into data mining. The authors' investigation revealed promising results in the stock market compared to popular machine learning and deep learning algorithms such as support vector machine (SVM), extreme gradient boosting (XGBoost), and long short-term memory (LSTM).

With the development of China's economy, the urban floating population is also increasing, resulting in a sharp increase in the amount of urban waste. How to recycle and dispose of municipal waste more efficiently has become the top concern of municipalities and other relevant departments. In this article, the above problem is transformed into the municipal waste collection vehicle routing problem (MWCVRP) to solve the problem with the minimum total waste transportation cost. Because the carrying capacity of different models is different, this article introduces a cost calculation criterion that combines the total mileage of different models of transport vehicles and the number of station services. A multi-model garbage truck path optimization model is established, and then a heuristic-based task dynamic assignment algorithm is designed to solve the problem. The Solomon dataset is used to verify the feasibility and effectiveness of the model and algorithm through experiments.

More and more graph embedding algorithms have been proposed, which makes the similarity judgment of graph structure more and more accurate. While exploring the similarity of neighborhood structures, the existence of weights should also be taken into account, so as to reflect the relational social network graph in the real world. We use Graphwave, a kind of algorithms for graph embedding with diffusion wavelets, to incorporate weight into numerical value to calculate, and to process the returned probability distribution parameters, so that we can get some analysis about the actual complex network. Our analysis can overcome the priori misjudgment problem based on the topological structure, and then obtain the actual similarity of the network structure from the results of graph embedding.