Group 1: DeepSeek Model and Factor Optimization - The DeepSeek model optimizes and generates stock selection factors based on the Alpha158 factor integrated with Qlib, covering five categories: intraday, volatility, price, volume, and price-volume correlation [3] - By optimizing these factors, DeepSeek significantly improves the predictive ability of most factors, particularly excelling in RankIC and ICIR metrics [3] - Generating new factors from scratch is challenging, but using existing optimized factor expressions as references allows for the generation of new high ICIR factors with fewer iterations [3] - In portfolio optimization, combining the optimized Alpha158 factors with newly generated factors using the Lasso model effectively enhances stock selection capabilities, especially in the full A-share data and the CSI 800 index combination [3] Group 2: New Incremental Policies and Economic Stimulus - On March 23, the Premier reaffirmed confidence in China's economic operation and macro policy implementation, emphasizing increased counter-cyclical adjustments and the potential introduction of new incremental policies, mainly referring to the possibility of central government leverage [4] - Following the National People's Congress, high-level leaders conducted intensive research in key regions such as Guizhou, Guangdong, and Fujian, focusing on technology, openness, foreign trade, employment, and security [4] - The State Council approved the "2025 Key Work Division Plan" on March 12, stressing the need to strengthen support for local governments, particularly in resource assurance, technological innovation, and reform and opening up, likely favoring major economic provinces like Zhejiang, Guangdong, and Fujian [4] - The central bank and other departments plan to stimulate the economy through policies such as reserve requirement ratio cuts, interest rate reductions, promoting foreign trade growth, and facilitating cross-border trade [4] - The "Special Action Plan to Boost Consumption" was issued on March 16, proposing measures to stimulate consumption through demand reform, supply upgrades, and livelihood guarantees, while also mentioning stabilizing the stock and real estate markets [4] Group 3: Hainan Huatie's Computing Power Leasing Business - Hainan Huatie (Huatie Dahuangfeng) has begun laying out its computing power leasing business and announced an investment plan for intelligent computing centers in 2024 [5] - As of December 2024, the company has signed computing power service orders totaling 2.475 billion yuan, and with a new contract of 3.69 billion yuan, the cumulative order amount has reached 6.165 billion yuan, 1.39 times its 2023 revenue [5] - The company’s former controlling shareholder has transferred its shares to Hainan State-owned Capital Holding Company, becoming the new controlling shareholder [5] - Expected net profits for the company from 2024 to 2026 are projected to be 619 million yuan, 846 million yuan, and 1.065 billion yuan, with EPS of 0.31 yuan, 0.43 yuan, and 0.54 yuan respectively [5] Group 4: Haiguang Information's Development in CPU and AI Computing - Haiguang Information, established in 2014, obtained AMD technology authorization in 2016 and initiated CPU product development, leveraging AMD's technology and its own R&D capabilities [7] - The company employs a "sell one generation, verify one generation, and develop one generation" strategy to promote continuous iteration of its Haiguang series CPUs and deep computing units (DCUs) [7] - Haiguang has significantly increased R&D investment, with product performance gradually catching up to global leaders while maintaining a leading position in the domestic market [7] - The Haiguang CPU is based on the x86 architecture, featuring a mature ecosystem and strong compatibility, particularly achieving seamless switching at the software and application levels [8] - Haiguang's DCU products, primarily used in supercomputing centers, have become leaders in domestic supercomputing chips, with ongoing development of the next-generation DCU, expected to significantly enhance AI performance [12] - The domestic information technology innovation industry is accelerating, with increasing domestic production rates supported by fiscal policies [10] - Haiguang is positioned to benefit from the rise of the domestic AI computing industry amid increasing U.S. chip export controls [12] Group 5: Investment Strategies During Earnings Season - Stocks disclosing last year's annual reports are suitable for shorter holding periods, while those disclosing quarterly reports are better for longer holding periods [13] - The excess return rate is typically better for shorter holding periods after annual report disclosures, while it improves with longer holding periods after quarterly report disclosures [13] - In the context of technology stocks experiencing box fluctuations, opportunities for low-cost entry in humanoid robots and DeepSeek + AI computing can be considered [13] - Investors should focus on stocks that perform well post-annual report release, keeping holding periods within T+5 for short-term gains [13] - After quarterly report disclosures, excess returns improve with extended holding periods, making it suitable to select stocks with sustained growth potential and consider holding until T+10 [13] - Maintaining dividend stocks as defensive assets is advisable, especially in a market lacking hotspots, with sectors like steel also considered for supplementary allocation [13]