{"paper_id":"2411.08382","title":"Hybrid Vector Auto Regression and Neural Network Model for Order Flow Imbalance Prediction in High Frequency Trading","authors":["Abdul Rahman","Neelesh Upadhye"],"paper_date":"2024-11-13","paper_type":"empirical","paper_url":"https://arxiv.org/pdf/2411.08382v1","page_url":"https://thequant.space/flowcharts/hybrid-vector-auto-regression-and-neural-network-model-for-o/","scores":{"empirical_rigor":7.5,"math_complexity":6,"quadrant":"Holy Grail"},"proposes":{"status":"available","statement":"The paper proposes a hybrid predictive model that combines Vector Auto Regression (VAR) with a simple feedforward neural network (FNN) to forecast Order Flow Imbalance (OFI) and assess trading intensity.","quote":"this paper introduces a hybrid predictive model that combines Vector Auto Regression (VAR) with a simple feedforward neural network (FNN) to forecast OFI and assess trading intensity.","summary":"In high frequency trading, accurate prediction of Order Flow Imbalance (OFI) is crucial for understanding market dynamics and maintaining liquidity. This paper introduces a hybrid predictive model that combines Vector Auto Regression (VAR) with a simple feedforward neural network (FNN) to forecast…","strategy":null},"assumptions":{"status":"available","items":[{"assumption":"The VAR component captures linear dependencies while the FNN models non‑linear residuals.","type":"model","stated":"explicit","quote":"The VAR component captures linear dependencies, while residuals are fed into the FNN to model non-linear patterns"},{"assumption":"Lag order p is selected using information criteria such as AIC or BIC.","type":"model","stated":"implied","quote":"past observations are considered, and is often determined through criteria such as the Akaike Information Criterion (AIC) or Bayesian Information Criterion (BIC)."},{"assumption":"The dataset is used without any normalization or resampling.","type":"data","stated":"explicit","quote":"gathered for a single cryptocurrency without applying any normalization or resampling."},{"assumption":"Synthetic data are generated to replicate characteristics of the real datasets.","type":"data","stated":"explicit","quote":"A synthetically generated dataset of 3,000 data points designed to replicate characteristics similar to the real datasets."},{"assumption":"Training uses early stopping to ensure model stability.","type":"model","stated":"explicit","quote":"Epochs: 50, with early stopping to ensure model stability."},{"assumption":"Trading intensity can be measured as a separate output from the hybrid model.","type":"market","stated":"explicit","quote":"Output Layer for Hybrid Model: Two neurons for predicting buy and sell orders individually, from which OFI is calculated."}]},"datasets":{"status":"available","items":[{"name":"Real Dataset 1 (BTCUSD)","provider":"Binance","kind":"crypto-onchain","access":"public","period":null,"universe":null,"quote":"Real Dataset 1 (BTCUSD): Contains approximately 3,000 data points representing BTCUSD transactions.","known_id":"binance"},{"name":"Real Dataset 2 (ETCUSDT)","provider":"Binance","kind":"crypto-onchain","access":"public","period":null,"universe":null,"quote":"Real Dataset 2 (ETCUSDT): Contains approximately 3,000 data points for ETCUSDT transactions.","known_id":"binance"},{"name":"Synthetic Dataset","provider":null,"kind":"synthetic","access":"unknown","period":null,"universe":null,"quote":"Synthetic Dataset: A synthetically generated dataset of 3,000 data points designed to replicate characteristics similar to the real datasets."}],"known":[{"id":"binance","name":"Binance","provider":"Binance","kind":"market-prices","access":"public","quote":"Using both synthetic and real trading data from Binance, we demonstrate that the hybrid model offers significant improvements in predictive accuracy and enhances strategic decision-making based on OFI dynamics.","mentions":4,"used_in":3}],"sample":{"start":null,"end":null,"market":null,"asset_class":null,"frequency":null},"data_resolution":"intraday-bars"},"transaction_costs":{"status":"not_applicable","modeled":"n/a","headline_net_of_costs":null,"detail":null,"quote":null,"source":"front_matter","rule":"R1","candidates":{"front_matter":"n/a","evidence_card":"unclear","evidence_card_quoted":false},"meaning":"the paper reports no trading or execution result"},"results":{"status":"available","evaluation_type":"forecast-accuracy","findings":[{"finding":"The hybrid VAR‑FNN model achieved lower MSE (0.00127) and higher R² (0.9946) than the FNN‑only model (MSE 0.00133, R² 0.9843) on training data.","kind":"performance","quote":"Hybrid VAR-FNN 0.00127 0.00541 0.9946"},{"finding":"On validation, the hybrid model outperformed VAR‑only and FNN‑only models with MSE 0.002, MAE 0.019, R² 0.997 and intensity accuracy 98.18% for BTCUSD.","kind":"predictive-accuracy","quote":"Hybrid VAR-FNN 0.002 0.019 0.997 98.18% 96.32%"},{"finding":"The hybrid model consistently achieved higher trading‑signal accuracy (≈96‑98%) compared to VAR‑only and FNN‑only models across real and synthetic datasets.","kind":"performance","quote":"The hybrid model achieved high accuracy in predicting trading signals for both real and synthetic datasets, showing significant improvements over the VAR-only and FNN-only models."}],"reported_performance":null,"out_of_sample":{"method":"unclear","headline_out_of_sample":null}},"independent_evidence":{"status":"none_found","supports":[],"contradicts":[],"qualifies":[],"compared_with":10,"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":"2026-11-14T17:54:23Z"},"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/"}