{"paper_id":"2307.00459","title":"Principal Component Analysis and Hidden Markov Model for Forecasting Stock Returns","authors":["Eugene W. Park"],"paper_date":"2023-07-02","paper_type":"empirical","paper_url":"https://arxiv.org/pdf/2307.00459v1","page_url":"https://thequant.space/flowcharts/principal-component-analysis-and-hidden-markov-model-for-for/","scores":{"empirical_rigor":6,"math_complexity":7.5,"quadrant":"Holy Grail"},"proposes":{"status":"available","statement":"The paper proposes a method that combines principal component analysis and hidden Markov models to forecast stock returns and build trading strategies.","quote":"This paper presents a method for predicting stock returns using principal component analysis (PCA) and the hidden Markov model (HMM) and tests the results of trading stocks based on this approach.","summary":"This paper presents a method for predicting stock returns using principal component analysis (PCA) and the hidden Markov model (HMM) and tests the results of trading stocks based on this approach. Principal component analysis is applied to the covariance matrix of stock returns for companies listed…","strategy":null},"assumptions":{"status":"available","items":[{"assumption":"Emission probabilities in each HMM state follow a Gaussian distribution.","type":"model","stated":"explicit","quote":"we assume the emission probability in each state in the state space, S, to have have a Gaussian distribution."},{"assumption":"The underlying Markov chain is equally likely to start in any state.","type":"model","stated":"explicit","quote":"we are assuming that it is equally likely for the underlying Markov chain to start in any state"},{"assumption":"State transition probabilities are equally likely between any pair of states.","type":"model","stated":"explicit","quote":"it is equally likely for the Markov chain to transit from one state to another"},{"assumption":"Trading can be executed at the exact closing price with no transaction costs.","type":"execution","stated":"explicit","quote":"we assume the following: 1. we are able to trade at exactly the closing time 2. we can exactly match the n assets with equal weight in our portfolio 3. and we do not incur any trading costs."},{"assumption":"Risk‑free rate is zero.","type":"model","stated":"explicit","quote":"We will assume the risk-free rate to be zero."},{"assumption":"Survivor bias is mitigated by requiring at least 400 return observations per period.","type":"data","stated":"implied","quote":"we are subject to survivor bias, but minimize this effect by ensuring that we have at least 400 data points of company returns for each time period."}]},"datasets":{"status":"available","items":[{"name":"Weekly returns of S&P 500 companies","provider":"finance.yahoo.com","kind":"market-prices","access":"public","period":"July 23, 2004 to July 8, 2016","universe":"S&P 500 constituents","quote":"We use data for the weekly returns of companies listed in the S&P500 to train our model. The returns are calculated using the closing stock prices, and each return sequence comprise the columns of our dataset, denoted X."}],"known":[],"sample":{"start":null,"end":null,"market":null,"asset_class":null,"frequency":null},"data_resolution":"weekly-monthly"},"transaction_costs":{"status":"available","modeled":"no","headline_net_of_costs":null,"detail":null,"quote":"we do not incur any trading costs.","source":"front_matter","rule":"R3","candidates":{"front_matter":"no","evidence_card":"no","evidence_card_quoted":true},"card_detail":null,"card_quote":"we do not incur any trading costs.","meaning":"a trading or execution result is reported without costs"},"results":{"status":"available","evaluation_type":"trading-pnl","findings":[{"finding":"Strategy 2 with p = 15% achieves an annualized Sharpe ratio of 1.36, more than 50% higher than the buy‑and‑hold benchmark.","kind":"performance","quote":"the strategy with the highest Sharpe ratio is strategy 2 with p = 15%. This outperforms"},{"finding":"Forecast accuracy (directional) is only slightly above 50%, with Strategy 1 yielding just above 50% winning probability.","kind":"predictive-accuracy","quote":"Strategy 1 clearly outperforms strategy 2 in forecasting the direction of the returns. Yet, strategy 1 yield just above 50%..."},{"finding":"Sharpe ratios are highly sensitive to the hyperparameter p, with a spread of up to 0.91 between the best and worst configurations.","kind":"statistical","quote":"ratios depend heavily on the hyperparameter p. The difference between the highest and lowest ratios of strategy is as high as 0.91."},{"finding":"The PCA+HMM model provides forecast accuracy slightly above 50% and can outperform the market when the right p is chosen.","kind":"performance","quote":"yields model-based trading strategies with forecast accuracy slightly above 50%."},{"finding":"Strategy 2 (p = 15%) shows lower cumulative return than buy‑and‑hold but is more stable, avoiding major downturns in late‑2014 and early‑2016.","kind":"performance","quote":"it rides out two ocassions of downturns, during late-2014 and early-2016, without much losses, contrary to the other strategies."}],"reported_performance":null,"out_of_sample":{"method":"walk-forward","headline_out_of_sample":null}},"independent_evidence":{"status":"none_found","supports":[],"contradicts":[],"qualifies":[],"compared_with":36,"citations":{"cited_by_in_corpus":0,"by_independent_authors":0,"how_cited":{"used_as_baseline":0,"uses_method_or_data":0,"extends":0,"discusses_findings":0,"mentions":0,"unclassified":0},"independent_papers":[],"scope":"Citations between papers in this corpus only; citations from outside it are not counted."}},"implementation":{"status":"available","public":false,"code_url":null,"note":"No repository presented by the authors as their own was found in the paper. Absence here is not proof that no code exists."},"replication":{"status":"none_found","card":null,"note":"No replication of this paper has been run by thequant.space yet."},"extraction":{"text_source":"fulltext","model":"openai/gpt-oss-120b","version":1,"extracted_at":"2026-11-14T18:25:20Z"},"provenance":{"claims":"Everything here is what the authors state in the paper. Nothing was re-run or reproduced by the platform.","quotes":"Assumptions, datasets, cost detail and findings were extracted by a language model. Every quote is verbatim paper text. Where an item has `support`, \"full\" means every figure in it appears in the quoted sentences and \"partial\" means a figure is stated elsewhere in the paper; items with a figure found nowhere in the paper were discarded.","datasets_known":"datasets.known lists well-known vendors and public datasets by a stable id, found by a dictionary over the full text where the sentence says the data is used; it is independent of the model and covers only names in the dictionary.","flags":"transaction_costs.modeled, results.evaluation_type and results.out_of_sample.method are platform classifications (structured-decision models); see https://thequant.space/score-guide/#penalty-flags","comparisons":"supports / contradicts / qualifies are a language model's reading of two papers' quoted findings, restricted to closely related papers that share no author. They are pointers to read, not verified replications, and coverage is partial.","implementation":"code_url is a repository the paper itself presents as its own (rule-scored, ambiguous cases judged by a model), checked for liveness when it was recorded."},"docs":"https://thequant.space/agents/"}