Quantitative Models and Construction Methods - **Model Name**: Competitive Advantage Portfolio **Construction Idea**: Focuses on analyzing industry competition barriers and identifying companies with unique management advantages in various industry categories such as "Shielded Barriers," "Intense Competition," "Steady Progress," and "Seeking Breakthroughs" [13][14] **Construction Process**: Combines "Shielded Barriers" industries with "Dominant + Cooperative Win-Win" companies and "Efficient Operations" companies in non-barrier industries to form the Competitive Advantage Portfolio [14] **Evaluation**: Provides a differentiated value quantification perspective compared to traditional factor investment [13] - **Model Name**: Safety Margin Portfolio **Construction Idea**: Emphasizes the gap between intrinsic value and market value, focusing on companies with sustainable competitive advantages and high ROIC [18] **Construction Process**: Calculates intrinsic value based on profitability metrics, selects the top 50 stocks with the highest safety margin from a competitive advantage stock pool, and weights them by dividend yield [18][20] **Evaluation**: Highlights the importance of intrinsic value estimation and sustainable profitability [18] - **Model Name**: Dividend Low Volatility Adjusted Portfolio **Construction Idea**: Avoids "high dividend traps" by focusing on sustainable profitability and excluding stocks with extreme price performance or abnormal debt ratios [25] **Construction Process**: Implements predictive models for dividend yield and applies negative screening criteria to optimize the portfolio [25] **Evaluation**: Addresses the risks of chasing high dividend yields without considering long-term value [25] - **Model Name**: AEG Valuation Potential Portfolio **Construction Idea**: Invests in companies with abnormal earnings growth (AEG) that exceed opportunity costs, focusing on undervalued growth potential [30][34] **Construction Process**: Uses the AEG_EP factor to select the top 100 stocks, then narrows down to the top 50 stocks with high dividend reinvestment/P ratios [34] **Evaluation**: Incorporates growth premiums into valuation models, providing a comprehensive perspective on future earnings potential [30][31] - **Model Name**: Cash Cow Portfolio **Construction Idea**: Evaluates companies based on free cash flow (FCF) and cash flow return (CFOR) to assess profitability and cash generation efficiency [38][40] **Construction Process**: Combines CFOR decomposition with ROE decomposition, focusing on high-quality stocks within the CSI 800 index [39][40] **Evaluation**: Enhances traditional DuPont analysis by integrating cash flow metrics for a more comprehensive evaluation [38] - **Model Name**: Large Model AI Stock Selection Portfolio **Construction Idea**: Utilizes FinLLM to process unstructured financial texts and integrates multi-dimensional validation methods such as chain-of-thought reasoning (COT), comparative analysis, and counterfactual reasoning [44][47] **Construction Process**: Applies FinLLM to extract signals from financial texts and uses a triangular validation system to ensure decision-making robustness [47][48] **Evaluation**: Overcomes limitations of traditional models by leveraging AI for non-structured data analysis and improving prediction accuracy [44][47] - **Model Name**: Governance Efficiency Portfolio **Construction Idea**: Analyzes MD&A disclosures to evaluate management transparency, financial consistency, and long-term value creation [53][54] **Construction Process**: Combines short-term profit guidance and financial consistency factors to form a base portfolio, then selects top 50 stocks using PB_ROE factor for valuation and profitability [57] **Evaluation**: Provides insights into management quality and strategic alignment, emphasizing governance as a key alpha source [53][57] --- Model Backtesting Results - **Competitive Advantage Portfolio**: Annualized return 20.41%, Sharpe ratio 0.93, IR 0.12, max drawdown -19.32%, Calmar ratio 1.06 [17] - **Safety Margin Portfolio**: Annualized return 20.27%, Sharpe ratio 1.02, IR 0.13, max drawdown -16.89%, Calmar ratio 1.20 [22] - **Dividend Low Volatility Adjusted Portfolio**: Annualized return 17.36%, Sharpe ratio 1.00, IR 0.15, max drawdown -21.61%, Calmar ratio 0.80 [26] - **AEG Valuation Potential Portfolio**: Annualized return 23.33%, Sharpe ratio 1.11, IR 0.16, max drawdown -24.04%, Calmar ratio 0.97 [36] - **Cash Cow Portfolio**: Annualized return 13.56%, Sharpe ratio 0.66, IR 0.13, max drawdown -19.80%, Calmar ratio 0.68 [42] - **Large Model AI Stock Selection Portfolio**: Annualized return 16.53%, Sharpe ratio 0.71, IR 0.17, max drawdown -33.01%, Calmar ratio 0.50 [49] - **Governance Efficiency Portfolio**: Annualized return 11.00%, Sharpe ratio 0.51, IR 0.23, max drawdown -23.74%, Calmar ratio 0.46 [59]

Quantitative Models and Construction 1. Model Name: Volume Model - **Model Construction Idea**: This model evaluates market timing based on trading volume dynamics[10][64] - **Model Evaluation**: The model currently signals a neutral stance for the short term[10][64] 2. Model Name: Low Volatility Model - **Model Construction Idea**: This model assesses market timing by analyzing low volatility trends in the market[10][64] - **Model Evaluation**: The model currently signals a neutral stance for the short term[10][64] 3. Model Name: Institutional Feature Model (Dragon-Tiger List) - **Model Construction Idea**: This model uses institutional trading features from the Dragon-Tiger list to predict market movements[10][64] - **Model Evaluation**: The model currently signals a bearish outlook for the short term[10][64] 4. Model Name: Feature Volume Model - **Model Construction Idea**: This model leverages specific volume features to predict market trends[10][64] - **Model Evaluation**: The model currently signals a bearish outlook for the short term[10][64] 5. Model Name: Intelligent CSI 300 Model - **Model Construction Idea**: This model applies intelligent algorithms to predict movements in the CSI 300 index[10][64] - **Model Evaluation**: The model currently signals a bullish outlook for the short term[10][64] 6. Model Name: Intelligent CSI 500 Model - **Model Construction Idea**: This model applies intelligent algorithms to predict movements in the CSI 500 index[10][64] - **Model Evaluation**: The model currently signals a bearish outlook for the short term[10][64] 7. Model Name: Limit-Up/Down Model - **Model Construction Idea**: This model evaluates market timing based on the frequency of limit-up and limit-down events[11][65] - **Model Evaluation**: The model currently signals a bullish outlook for the mid-term[11][65] 8. Model Name: Calendar Effect Model - **Model Construction Idea**: This model incorporates calendar-based patterns to predict market movements[11][65] - **Model Evaluation**: The model currently signals a neutral stance for the mid-term[11][65] 9. Model Name: Long-Term Momentum Model - **Model Construction Idea**: This model evaluates long-term market trends using momentum indicators[12][66] - **Model Evaluation**: The model currently signals a neutral stance across all broad-based indices for the long term[12][66] 10. Model Name: A-Share Comprehensive Weapon V3 Model - **Model Construction Idea**: This model integrates multiple signals to provide a comprehensive market timing prediction[13][67] - **Model Evaluation**: The model currently signals a bearish outlook for the A-share market[13][67] 11. Model Name: A-Share Comprehensive CSI 2000 Model - **Model Construction Idea**: This model focuses on the CSI 2000 index, combining various timing signals[13][67] - **Model Evaluation**: The model currently signals a neutral stance for the A-share market[13][67] 12. Model Name: Turnover-to-Volatility Model (Hong Kong Market) - **Model Construction Idea**: This model evaluates market timing in the Hong Kong market by analyzing turnover relative to volatility[14][68] - **Model Evaluation**: The model currently signals a bullish outlook for the mid-term[14][68] --- Model Backtesting Results 1. Volume Model - **Short-Term Signal**: Neutral[10][64] 2. Low Volatility Model - **Short-Term Signal**: Neutral[10][64] 3. Institutional Feature Model (Dragon-Tiger List) - **Short-Term Signal**: Bearish[10][64] 4. Feature Volume Model - **Short-Term Signal**: Bearish[10][64] 5. Intelligent CSI 300 Model - **Short-Term Signal**: Bullish[10][64] 6. Intelligent CSI 500 Model - **Short-Term Signal**: Bearish[10][64] 7. Limit-Up/Down Model - **Mid-Term Signal**: Bullish[11][65] 8. Calendar Effect Model - **Mid-Term Signal**: Neutral[11][65] 9. Long-Term Momentum Model - **Long-Term Signal**: Neutral across all broad-based indices[12][66] 10. A-Share Comprehensive Weapon V3 Model - **Comprehensive Signal**: Bearish[13][67] 11. A-Share Comprehensive CSI 2000 Model - **Comprehensive Signal**: Neutral[13][67] 12. Turnover-to-Volatility Model (Hong Kong Market) - **Mid-Term Signal**: Bullish[14][68]

Core Viewpoint - The article emphasizes the growing popularity and effectiveness of quantitative stock selection strategies in the A-share market, particularly in a challenging market environment where traditional index-based returns are hard to achieve [2]. Group 1: Market Context - The A-share market has been characterized by a lack of index opportunities, often experiencing fluctuations or declines, making it difficult to obtain beta returns [2]. - In this context, investors are increasingly relying on alpha returns, leading to a surge in interest for quantitative stock selection products [2]. Group 2: Advantages of Quantitative Stock Selection - Quantitative stock selection offers several advantages over traditional subjective investment strategies: 1. Objective decision-making that mitigates human emotional biases such as fear and greed [2]. 2. Efficient scanning of the entire market to identify overlooked investment opportunities [2]. 3. Rapid response and discipline in executing trades, ensuring timely actions based on market signals [2]. Group 3: Performance Metrics - As of May 30, 2025, quantitative stock selection strategies have outperformed major A-share indices over various time frames: - 6-month return: 7.24% - 1-year return: 26.57% - 3-year return: 36.88% - 5-year return: 86.17% [3]. Group 4: Company Profile - 微观博易 - 微观博易 has demonstrated strong performance in the quantitative stock selection space, with its product "微观博易-夏之" achieving significant returns since transitioning to a quantitative strategy [4]. - The company employs a multi-factor stock selection model that combines both human and machine-generated factors, aiming for a balanced and diversified portfolio [5]. - The firm has invested heavily in infrastructure, including proprietary technology for low-latency trading, which enhances its competitive edge in the market [10]. Group 5: Team and Strategy - 微观博易's team consists of over 30 professionals with backgrounds from prestigious institutions and experience in top-tier trading firms [7][9]. - The company focuses on a multi-asset strategy, with a strong emphasis on risk management and maximizing risk-adjusted returns [9][10]. - The firm plans to expand its quantitative stock selection product line, indicating a commitment to growth and innovation in this area [13].

Group 1 - The fund name is "招商红利量化选股混合型证券投资基金" (招商红利量化选股混合) with A and C share classes [25][26] - The fund type is a mixed securities investment fund and operates as a contractual open-end fund [2][26] - The fund has a maximum fundraising limit of 3 billion RMB (excluding interest during the fundraising period) [3][30] Group 2 - The fundraising period is from May 14, 2025, to May 28, 2025, with the possibility of adjustments based on subscription conditions [2][33] - The fund is open to individual investors, institutional investors, qualified foreign investors, and other investors permitted by laws and regulations [6][32] - The minimum total fundraising amount is set at 200 million shares [29] Group 3 - The fund's A share class will charge subscription fees, while the C share class will not [36][37] - The initial subscription price for each fund share is set at 1.00 RMB [28][36] - Investors can subscribe multiple times during the fundraising period, with specific minimum amounts depending on the sales channel [5][42] Group 4 - The fund management company is 招商基金管理有限公司 (招商基金) and the custodian is 中信银行股份有限公司 (CITIC Bank) [1][2] - The fund will utilize a quantitative model for stock selection, which may involve risks related to model effectiveness and market conditions [10][11] - The fund's investment strategy includes a focus on stocks, with a target allocation of 60%-95% of its assets in stocks and depositary receipts [10]

Group 1 - The core viewpoint of the news is that Jing Sheng Machinery continues to implement its development strategy of "advanced materials and advanced equipment," achieving significant revenue and profit growth [1] - For the reporting period, Jing Sheng Machinery reported operating revenue of 1,757.66 million yuan and a net profit of 250.97 million yuan, indicating strong financial performance [1] - The company is accelerating the domestic substitution process in the semiconductor equipment market, successfully developing various 12-inch semiconductor equipment that meet international advanced standards [1] Group 2 - Jing Sheng Machinery is rapidly advancing its 8-inch silicon carbide substrate production capacity, with significant market expansion results [1] - The quartz crucible business has improved production efficiency by creating a fully automated production platform, positioning Jing Hong Precision as a core component supplier [1] - Major clients include well-known listed companies and large enterprises such as TCL Zhonghuan and Changdian Technology, with good payment performance on existing orders [1] Group 3 - The company aims to maintain its development strategy of "advanced materials and advanced equipment" while building a platform company that promotes diversified business collaboration [1]

Core Viewpoint - Hongda Co., Ltd. has reported a turnaround in net profit for 2024, but its stock price has significantly retraced since reaching a peak of 9.30 yuan per share in early November 2024, nearly erasing the gains made in October 2024 [1] Group 1: Financial Performance - In 2024, Hongda Co., Ltd. achieved total operating revenue of 3.409 billion yuan, a year-on-year increase of 12.68% [7] - The net profit attributable to shareholders was 36.11 million yuan, recovering from a loss of 95.84 million yuan in the previous year [7] - The company reported a non-recurring net profit of 13.12 million yuan, compared to a loss of 104.63 million yuan in the previous year [7] - Despite the recovery, the net profit level in 2024 has not fully returned to the 2022 level, where it was 60.16 million yuan [7] Group 2: Institutional Investment - By the end of 2024, institutional investors held a total of 9.57 billion shares of Hongda, an increase from 8.42 billion shares at the end of Q3 2024, with 44 new funds and 1 insurance company entering as investors [2] - The largest institutional holders include a single insurance company holding 71.02 million shares and 45 funds holding a combined 68.91 million shares [2] - Notable funds include the China Europe Industry Outlook Mixed Securities Investment Fund, which newly acquired 10 million shares, ranking as the ninth largest holding [4] Group 3: Shareholder Changes - The original controlling shareholder, Sichuan Hongda Industrial Co., Ltd., entered bankruptcy reorganization in June 2023, with the reorganization plan approved in July 2024 [6] - Following the reorganization, the controlling shareholder changed to Shudao Investment Group, which now holds 26.39% of the total shares [6] - As of October 31, 2024, Shudao Group and its concerted parties controlled 31.31% of Hongda's total shares [6] Group 4: Historical Issues and Debt - Hongda has faced significant debt due to a major lawsuit involving Yunnan Jinding Zinc Industry Co., Ltd., which resulted in a repayment of 650.67 million yuan, with remaining obligations totaling 423.43 million yuan [8] - The company has resolved its historical issues and is currently in the process of a stock issuance to raise 2.853 billion yuan for debt repayment and working capital [9]

Core Viewpoint - The article discusses the performance and strategies of various "Fixed Income +" funds in the current market environment, highlighting their unique risk-return characteristics and providing insights for investors to consider different options for asset allocation [2][15]. Group 1: Fund Performance and Strategies - The article highlights four funds with different strategies: 1. Huaan Enhanced Income Bond focuses on "Fixed Income + Convertible Bonds" 2. Huatai-PB Dingli Flexible Allocation Mixed focuses on "Fixed Income + Cyclical Stocks" (mainly gold stocks) 3. E Fund Rui Jin Mixed focuses on "Fixed Income + Dividends" 4. China Merchants Anyang Bond focuses on "Fixed Income + Quantitative Stock Selection" [2]. - Huaan Enhanced Income Bond, managed by Zheng Weishan, achieved a 1-year return of 11.41%, significantly outperforming its benchmark of 5.86%, ranking in the top 2% of its category [3][4]. - Huatai-PB Dingli Flexible Allocation Mixed, managed by Zheng Qing and Dong Chen, reported a 1-year return of 6.72%, with a flexible bond portfolio primarily consisting of financial bonds [7][8]. - E Fund Rui Jin Mixed, managed by Yang Kang, achieved a 1-year return of 10.90%, focusing on high-dividend assets and actively adjusting stock positions based on market conditions [10][11]. - China Merchants Anyang Bond, managed by Yin Xiaohong and Cai Zhen, reported a 1-year return of 9.72%, utilizing a "Fixed Income + Quantitative" strategy to balance risk and return [12][14]. Group 2: Investment Strategies and Asset Allocation - Huaan Enhanced Income Bond has shifted its bond holdings to focus on convertible bonds, maintaining over 70% in this asset class since 2023, which provides both defensive and offensive characteristics [3][5]. - Huatai-PB Dingli's bond holdings are primarily in financial bonds, with a flexible allocation strategy that adjusts based on market conditions, aiming to capture opportunities in a declining interest rate environment [7][8]. - E Fund Rui Jin Mixed employs a strategy focused on high-dividend stocks, adjusting its equity positions based on specific market indicators to optimize returns while managing risk [10][11]. - China Merchants Anyang Bond utilizes a quantitative approach for stock selection, focusing on industry rotation and maintaining a balance between different sectors to adapt to market changes [13][14]. Group 3: Conclusion and Investor Considerations - The article emphasizes the importance of understanding the underlying strategies of "Fixed Income +" funds, as different approaches can lead to varying performance outcomes, allowing investors to find suitable products based on their risk preferences and investment goals [15].

广发证券首席金工分析师 安宁宁 anningning@gf.com.cn 广发证券资深金工分析师 陈原文 chenyuanwen@gf.com.cn 联系人：广发证券金工研究员 林涛 gflintao@gf.com.cn 广发金工安宁宁陈原文团队 摘要 神经常微分方程： 在机器学习国际顶会NeurIPS 2018上，Chen等人发表的论文《Neural Ordinary Differential Equations》获得了大会的最佳论文奖。简单来 说，一个常见的ResNet网络通常由多个形如h_{t+1}=f(h_t,_t)+h_t的残差结构所组成。在常规求解中，需计算出每一个残差结构中最能拟合训练数据的网 络参数。而该论文提出，假设当ResNet网络中的残差结构无限堆叠时，则每一个残差结构的参数都可以通过求解同一个常微分方程来获得。 液态神经网络： 基于上述工作，来自麻省理工学院的Ramin Hasani等人，创新性地以常微分方程的形式描述循环神经网络的隐藏状态变化，提出了一类被 称之为液态神经网络的模型，这些研究成果被发表在《Nature：Machine Intelligence》等国际顶级期刊上。此类模 ...