Figure 6 - uploaded by Arturo Cifuentes
Content may be subject to copyright.
Pair-plot comparison of synthetic versus original data average KS-test across all dimensions, divided by cluster. Values are scaled by the maximum KS-test score of each row.
Context in source publication
Context 1
... allowed us to compare the similarities between samples where the original and synthetic data originated from the same cluster versus those from different clusters. The results of this exercise, displayed in figure 6, reveal that the synthetic data conditioned on the same cluster as the original data typically yielded the highest KS-test scores compared to data generated from other clusters. This finding provides further evidence of the effectiveness of the cluster-based approach to produce synthetic data that replicates not only the broad characteristics of the original dataset but also all the key elements of all the different market regimes. ...
Similar publications
China faces a structural-demographic crisis characterized by low fertility rates, an aging population, and ecological imbalances. Despite policy interventions, fertility intentions remain a concern. This study explores the complex relationship between household financial asset allocation and fertility intentions using CFPS data. The findings reveal...
As an important asset class, the allocation decision of real estate is of great significance to the wealth accumulation of individuals and families. This paper builds a theoretical framework based on the theoretical perspective of human capital, takes the CHFS database in 2015,2017 and 2019 as the benchmark, and constructs the Probit model, two-way...
This study examines the impact of interest rates and inflation during the 2019-2023 period on asset allocation strategies in Indonesia's investment portfolio. Investment decisions are heavily influenced by two macroeconomic components, namely interest rates and inflation. These factors mainly affect how assets comprise a portfolio. In this study, d...
We build a state-of-the-art dynamic model of private asset allocation that considers five key features of private asset markets: (1) the illiquid nature of private assets, (2) timing lags between capital commitments, capital calls, and eventual distributions, (3) time-varying business cycle conditions, (4) serial correlation in observed private ass...
Research has shown that investment success is largely driven by asset allocation. With the dramatic growth in number of exchange-traded funds (ETFs), individual investors have gained access to a wicie variety of funds including funds representing non-traditional asset classes, This proliferation of ETFs allows investors to take advantage of high re...
Citations
... In terms of deep probabilistic modeling of financial time series, prior work has applied deep learning approaches to model a single financial time series (e.g., [4,31]), or multivariate time series (e.g., [21,25,27,28,32]). Of this work, Tepelyan and Gopal [28] (BDG 1 ) were the first to scale up to hundreds of stocks through combining machine learning with factor modeling, specifically Fama-French factor modeling. ...
... Mean and Covariance. Unlike prior work that has worked on multivariate generative modeling (e.g., [21,25,27,28,32])), our model is able to compute the mean ( + ) and the covariance matrix without sampling (Σ + Σ ) leading to significant speedup during inference. ...
The use of machine learning for statistical modeling (and thus, generative modeling) has grown in popularity with the proliferation of time series models, text-to-image models, and especially large language models. Fundamentally, the goal of classical factor modeling is statistical modeling of stock returns, and in this work, we explore using deep generative modeling to enhance classical factor models. Prior work has explored the use of deep generative models in order to model hundreds of stocks, leading to accurate risk forecasting and alpha portfolio construction; however, that specific model does not allow for easy factor modeling interpretation in that the factor exposures cannot be deduced. In this work, we introduce NeuralFactors, a novel machine-learning based approach to factor analysis where a neural network outputs factor exposures and factor returns, trained using the same methodology as variational autoencoders. We show that this model outperforms prior approaches both in terms of log-likelihood performance and computational efficiency. Further, we show that this method is competitive to prior work in generating realistic synthetic data, covariance estimation, risk analysis (e.g., value at risk, or VaR, of portfolios), and portfolio optimization. Finally, due to the connection to classical factor analysis, we analyze how the factors our model learns cluster together and show that the factor exposures could be used for embedding stocks.
