LUCIDA:如何利用多因子策略构建强大的加密资产投资组合(因子正交化篇)

Odaily星球日报Published on 2024-02-08Last updated on 2024-02-08

Abstract

在进行大类因子合成前,需要进行因子正交化来消除共线性,本篇就三种正交方法进行展开。

书接上回,关于《用多因子模型构建强大的加密资产投资组合》系列文章中,我们已经发布了四篇:《理论基础篇》《数据预处理篇》《因子有效性检验篇》《大类因子分析:因子合成篇》

在上一篇中,我们具体解释了因子共线性(因子之间相关性较高)的问题,在进行大类因子合成前,需要进行因子正交化来消除共线性。

LUCIDA:如何利用多因子策略构建强大的加密资产投资组合(因子正交化篇)

一、因子正交化的数学推导

从多因子截面回归角度,建立因子正交化体系。

LUCIDA:如何利用多因子策略构建强大的加密资产投资组合(因子正交化篇)

LUCIDA:如何利用多因子策略构建强大的加密资产投资组合(因子正交化篇)

LUCIDA:如何利用多因子策略构建强大的加密资产投资组合(因子正交化篇)

LUCIDA:如何利用多因子策略构建强大的加密资产投资组合(因子正交化篇)

LUCIDA:如何利用多因子策略构建强大的加密资产投资组合(因子正交化篇)

所以,

LUCIDA:如何利用多因子策略构建强大的加密资产投资组合(因子正交化篇)

LUCIDA:如何利用多因子策略构建强大的加密资产投资组合(因子正交化篇)

LUCIDA:如何利用多因子策略构建强大的加密资产投资组合(因子正交化篇)

LUCIDA:如何利用多因子策略构建强大的加密资产投资组合(因子正交化篇)

LUCIDA:如何利用多因子策略构建强大的加密资产投资组合(因子正交化篇)

LUCIDA:如何利用多因子策略构建强大的加密资产投资组合(因子正交化篇)

LUCIDA:如何利用多因子策略构建强大的加密资产投资组合(因子正交化篇)

LUCIDA:如何利用多因子策略构建强大的加密资产投资组合(因子正交化篇)

二、三种正交方法的具体实现

1.施密特正交

LUCIDA:如何利用多因子策略构建强大的加密资产投资组合(因子正交化篇)

LUCIDA:如何利用多因子策略构建强大的加密资产投资组合(因子正交化篇)

LUCIDA:如何利用多因子策略构建强大的加密资产投资组合(因子正交化篇)

LUCIDA:如何利用多因子策略构建强大的加密资产投资组合(因子正交化篇)

施密特正交是一种顺序正交方法,因此需要确定因子正交的顺序,常见的正交顺序有固定顺序(不同截面上取同样的正交次序),以及动态顺序(在每个截面上根据一定规则确定其正交次序)。施密特正交法的优点是按同样顺序正交的因子有显式的对应关系,但是正交顺序没有统一的选择标准,正交后的表现可能受到正交顺序标准和窗口期参数的影响。

LUCIDA:如何利用多因子策略构建强大的加密资产投资组合(因子正交化篇)

2.规范正交

LUCIDA:如何利用多因子策略构建强大的加密资产投资组合(因子正交化篇)

LUCIDA:如何利用多因子策略构建强大的加密资产投资组合(因子正交化篇)# 规范正交 def Canonical(self):
 overlapping_matrix = (time_tag_data.shape[ 1 ] - 1) * np.cov(time_tag_data.astype(float))
 # 获取特征值和特征向量
 eigenvalue, eigenvector = np.linalg.eig(overlapping_matrix)
 # 转换为 np 中的矩阵
 eigenvector = np.mat(eigenvector)
 transition_matrix = np.dot(eigenvector, np.mat(np.diag(eigenvalue ** (-0.5))))
 orthogonalization = np.dot(time_tag_data.T.values, transition_matrix)
 orthogonalization_df = pd.DataFrame(orthogonalization.T, index = pd.MultiIndex.from_product([time_tag_data.index, [time_tag]]), columns=time_tag_data.columns)
 self.factor_orthogonalization_data = self.factor_orthogonalization_data.append(orthogonalization_df)

3.对称正交

施密特正交由于在过去若干个截面上都取同样的因子正交顺序,因此正交后的因子和原始因子有显式的对应关系,而规范正交在每个截面上选取的主成分方向可能不一致,导致正交前后的因子没有稳定的对应关系。由此可见,正交后组合的效果,很大一部分取决于正交前后因子是否有稳定的对应关系。

对称正交尽可能的减少对原始因子矩阵的修改而得到一组正交基。这样能够最大程度地保持正交后因子和原因子的相似性。并且避免像施密特正交法中偏向正交顺序中靠前的因子。

LUCIDA:如何利用多因子策略构建强大的加密资产投资组合(因子正交化篇)

对称正交的性质:

  1. 与施密特正交相比,对称正交不需要提供正交次序,对每个因子是平等看待的

  2. 在所有正交过渡矩阵中,对称正交后的矩阵和原始矩阵的相似性最大,即正交前后矩阵的距离最小。

LUCIDA:如何利用多因子策略构建强大的加密资产投资组合(因子正交化篇)

Related Reads

The Reality of Payments in Latin America Is Not What You Think

The payment landscape in Latin America is undergoing a fundamental shift, driven by on-the-ground realities that challenge common perceptions. Based on over 500 hours of field research across the region, key insights emerge. Firstly, QR code payments, like Brazil's Pix, are becoming the dominant payment method in most emerging markets, overtaking cards. However, these domestic instant payment systems lack international interoperability, creating a significant gap for cross-border users. Secondly, the narrative around crypto cards is often misunderstood; their primary volume comes from high-net-worth professionals using them for salary conversions (e.g., USDT to local currency via Pix), not retail micro-payments. Competition in payments is shifting from customer acquisition to controlling the settlement layer, leading fintechs to acquire banking licenses for efficiency. Thirdly, treating "Latin America" as a single market is a mistake. Countries like Argentina, Brazil, and Mexico have distinct economic realities, user segments, and regulatory approaches. Brazil alone has at least five distinct user segments with different financial flows. Overlooked markets like Guatemala, Honduras, and El Salvador (the "forgotten five") offer high remittance volumes with lower competitive density. Finally, regulation in Latin America is often ahead of the US, with clearer frameworks for digital assets and a pragmatic approach from regulators focused on safety rather than obstruction. The margin on stablecoin forex is rapidly compressing toward zero, meaning future winners will be those building value-added services on top of the infrastructure, not just the cheapest exchange.

marsbitJust now

The Reality of Payments in Latin America Is Not What You Think

marsbitJust now

Making Music in a Bear Market: The Survival Experiment of a Bitcoin Band

"Orange Pill Jam: A Bitcoin Band's Survival in the Bear Market" Orange Pill Jam is a musical group exploring themes of financial sovereignty and privacy, born from the Bitcoin community. Formed after singer Mermaid performed her song "Dollar Apocalypse" at a 2022 conference, the band creates music intended for both Bitcoin enthusiasts and general audiences. Their creative process involves Mermaid writing lyrics and melodies, which producer/multi-instrumentalist Michi then shapes with a precise, rhythm-focused approach, often demanding numerous retakes to achieve his unique standard of timing. Their songs, like "Cypherpunks' Manifesto" and "Fire of Freedom," tackle concepts of digital privacy, the pitfalls of "free" services, and personal sovereignty, influenced by experiences in places like El Salvador. Despite operating in a crypto bear market with a Copyleft model (offering music for free sharing/remixing and accepting optional Bitcoin donations), they face practical challenges. Their growth is slow on platforms like YouTube and Spotify, which aren't optimized for their niche content. The band also navigates the rise of AI-generated music. While acknowledging AI's efficiency for certain tasks, they believe human creativity occupies a unique space that algorithms cannot replicate—the ability to create new genres and capture intangible rhythmic feeling. For Orange Pill Jam, the core argument for both Bitcoin in a downturn and human artistry in the AI age lies in this irreplaceable, intentional, and imperfectly human creative process. Their project persists as an anti-algorithm experiment, valuing the unquantifiable impact of music over scalable metrics.

marsbit7m ago

Making Music in a Bear Market: The Survival Experiment of a Bitcoin Band

marsbit7m ago

Latin America's Payments Landscape Is Not What You Think It Is

This report challenges common misconceptions about Latin America's payment landscape, based on over 500 hours of firsthand research. Key findings include: 1) Crypto card transaction volume primarily comes from high-net-worth individuals receiving USDT salaries, not retail spending. 2) QR code payments (e.g., Brazil's Pix, Argentina's Mercado Pago) are the dominant payment method across most emerging markets, not cards. 3) A major untapped opportunity lies in enabling cross-border interoperability between domestic instant payment systems. 4) Payment competition is shifting from customer acquisition to owning the settlement layer (e.g., acquiring banks). 5) Latin America is not a single market; Brazil, Mexico, Argentina, and smaller "forgotten five" countries (e.g., Guatemala, Honduras) have vastly different dynamics. 6) Stablecoin-to-fiat conversion margins are collapsing toward zero, pushing companies to build value-added services on top. 7) Future payment winners will be multi-country brands, not single-corridor specialists. 8) Marketing must target specific user segments (e.g., digital nomads, unbanked immigrants) with tailored messaging, not a generic "Brazilian" audience. 9) Contrary to perception, Latin American regulators are often ahead of the US in creating frameworks for digital assets and instant payments, with clear licensing deadlines. The core takeaway is that the region's payment rules are being rewritten, moving beyond cards and stablecoin arbitrage towards integrated, cross-border QR-based solutions.

链捕手17m ago

Latin America's Payments Landscape Is Not What You Think It Is

链捕手17m ago

Trading

Spot
Futures
活动图片