{"paper_id":"2411.12753","title":"Supervised Autoencoders with Fractionally Differentiated Features and Triple Barrier Labelling Enhance Predictions on Noisy Data","authors":["Bartosz Bieganowski","Robert Ślepaczuk"],"paper_date":"2024-11-06","paper_type":"empirical","paper_url":"https://arxiv.org/pdf/2411.12753v2","page_url":"https://thequant.space/flowcharts/supervised-autoencoders-with-fractionally-differentiated-fea/","scores":{"empirical_rigor":6,"math_complexity":7.5,"quadrant":"Holy Grail"},"proposes":{"status":"available","statement":"The paper proposes a supervised autoencoder‑MLP model with fractionally differentiated features and triple‑barrier labeling to improve cryptocurrency trading performance.","quote":"This paper investigates the enhancement of financial time series forecasting with the use of neural networks through supervised autoencoders (SAE), to improve investment strategy performance.","summary":"This paper investigates the enhancement of financial time series forecasting with the use of neural networks through supervised autoencoders (SAE), to improve investment strategy performance. Using the Sharpe and Information Ratios, it specifically examines the impact of noise augmentation and…","strategy":"SAE-MLP triple barrier trading strategy"},"assumptions":{"status":"available","items":[{"assumption":"Orders are assumed to have no market impact and are executed instantly at the last close price.","type":"execution","stated":"explicit","quote":"For simplicity, we assume that the orders we place will not have an effect on the market and that they are executed instantly, at the last close price."},{"assumption":"Transaction costs are a flat 0.05% per opening or closing of a position.","type":"execution","stated":"explicit","quote":"Transaction cost for all assets amounts to 0.05%."},{"assumption":"Pre‑2020 data are used exclusively for training.","type":"data","stated":"explicit","quote":"The data spans from January 2016 to April 2022, with pre‑2020 data used exclusively for training."},{"assumption":"Fractional differentiation order d is chosen as the minimum coefficient that makes the series stationary (ADF test).","type":"model","stated":"explicit","quote":"that makes the fractionally differentiated series stationary (i.e., passes the Augmented Dickey‑Fuller test)."},{"assumption":"Stop‑losses and take‑profits from triple barrier labeling are executed immediately and perfectly.","type":"execution","stated":"explicit","quote":"assumes that stop‑losses and take profits from triple barrier labeling are executed immediately and perfectly."},{"assumption":"Walk‑forward validation is assumed to reflect real‑world model updating and retraining.","type":"statistical","stated":"implied","quote":"Walk‑forward validation provides a more accurate estimate of the model’s performance in real‑world scenarios, where models need to be updated and retrained regularly."}]},"datasets":{"status":"available","items":[{"name":"Binance 1‑minute OHLC bars for Bitcoin, Ethereum, Litecoin","provider":"Binance","kind":"market-prices","access":"public","period":"January 2016 to April 2022","universe":"BTC, ETH, LTC","quote":"Each of the time series was obtained from Binance in open high low close (OHLC) format with a 1‑minute bar frequency.","known_id":"binance"},{"name":"FirstRateData macro‑economic feature series","provider":"FirstRateData","kind":"macro","access":"unknown","period":null,"universe":null,"quote":"Feature time series, sourced from FirstRateData include: • Initial Claims (ICSA): Unemployment data from FRED. • WTI Crude Oil futures... •"}],"known":[{"id":"binance","name":"Binance","provider":"Binance","kind":"market-prices","access":"public","quote":"Each of the time series was obtained from Binance in open high low close (OHLC) format with a 1-minute bar frequency.","mentions":1,"used_in":1},{"id":"fred","name":"FRED","provider":"Federal Reserve Bank of St. Louis","kind":"macro","access":"public","quote":"Feature time series, sourced from FirstRateData include: • Initial Claims (ICSA): Unemployment data from FRED. • WTI Crude Oil futures: Indicator of economic growth and energy demand. • Henry Hub Natural Gas futures: Another energy demand indicator. • Corn futures: Reflects food-related economic factors. arXiv:2411.12753v2 [q-fin.TR] 30 Nov 2024 ICAIF-24, November 14-16, 2024, New York City,","mentions":1,"used_in":1}],"sample":{"start":"2016-01","end":"2022-04","market":"Bitcoin, Litecoin, Ethereum","asset_class":"crypto","frequency":"intraday"},"data_resolution":"intraday-bars"},"transaction_costs":{"status":"available","modeled":"unclear","headline_net_of_costs":false,"detail":null,"quote":null,"source":"conflict","rule":"R4b","candidates":{"front_matter":"partial","evidence_card":"yes","evidence_card_quoted":true},"conflict_kind":"graded","card_detail":"A flat 0.05% cost is applied to each opening and closing of a position.","card_quote":"Transaction cost for all assets amounts to 0.05%.","meaning":"the platform classifier and the paper-text extraction disagree, so no value is asserted"},"results":{"status":"available","evaluation_type":"trading-pnl","findings":[{"finding":"For Litecoin, the 30‑minute SAE strategy achieved the highest total return (382.49%) and annualized return (59.37%), outperforming buy‑and‑hold (118.92% total return).","kind":"performance","quote":"the most favorable risk‑adjusted metrics, with the highest information ratio (2.11) and starred information ratio (4.85), coupled with the lowest annualized standard deviation (28.09%) and maximum drawdown (25.88%)."},{"finding":"Across an equally weighted portfolio of BTC, ETH, and LTC, the 30‑minute SAE strategy achieved the highest information ratio (2.56) and starred information ratio (7.63) with the lowest annualized standard deviation (27.37%) and maximum drawdown (23.55%).","kind":"performance","quote":"The 30‑minute strategy demonstrated superior risk‑adjusted performance with the highest information ratio (2.56) and starred information ratio (7.63), coupled with the lowest annualized standard deviation (27.37%) and maximum drawdown (23.55%)."},{"finding":"All SAE strategies outperformed the buy‑and‑hold approach in risk‑adjusted metrics, even when returns were lower.","kind":"performance","quote":"Notably, all SAE strategies outperformed the buy‑and‑hold approach in terms of risk‑adjusted metrics, with even the 10‑minute strategy showing improved risk characteristics despite lower returns."}],"reported_performance":{"strategy":"SAE-MLP triple barrier trading strategy","sharpe":null,"sharpe_benchmark":null,"annual_return_pct":null,"total_return_pct":382.49,"max_drawdown_pct":25.88},"out_of_sample":{"method":"walk-forward","headline_out_of_sample":true}},"independent_evidence":{"status":"none_found","supports":[],"contradicts":[],"qualifies":[],"compared_with":21,"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":"2026-11-14T17:55:33Z"},"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/"}