{"paper_id":"2406.19399","title":"Predicting Customer Goals in Financial Institution Services: A Data-Driven LSTM Approach","authors":["Andrew Estornell","Stylianos Loukas Vasileiou","William Yeoh","Daniel Borrajo","Rui Silva"],"paper_date":"2024-05-22","paper_type":"empirical","paper_url":"https://arxiv.org/pdf/2406.19399v1","page_url":"https://thequant.space/flowcharts/predicting-customer-goals-in-financial-institution-services/","scores":{"empirical_rigor":6,"math_complexity":2.5,"quadrant":"Street Traders"},"proposes":{"status":"available","statement":"The paper proposes two models for predicting customer goals and actions: a standard LSTM and an LSTM enhanced with state-space graph embeddings.","quote":"we use historical customer traces generated by a realistic simulator and present two simple models for predicting customer goals and future actions -- an LSTM model and an LSTM model enhanced with state-space graph embeddings.","summary":"In today’s competitive financial landscape, understanding and anticipating customer goals is crucial for institutions to deliver a personalized and optimized user experience. This has given rise to the problem of accurately predicting customer goals and actions. Focusing on that problem, we use…","strategy":null},"assumptions":{"status":"available","items":[{"assumption":"The synthetic simulator generates data that accurately reflects real banking customer behavior.","type":"data","stated":"explicit","quote":"By leveraging this simulator, they are able to create datasets that accurately represent the intricacies and dynamics of customer traces in a banking environment"},{"assumption":"A frequency threshold of 10 occurrences is sufficient to define meaningful nodes and edges in the state graph.","type":"model","stated":"explicit","quote":"The graph was constructed by examining all customer trajectories in the training dataset and defining nodes and edges as any state or transition action which appeared in more than 10 times. We found 10 to be the best frequency threshold when defining the graph"},{"assumption":"Customer sequences are represented using bag‑of‑words and one‑hot encodings.","type":"model","stated":"explicit","quote":"Specifically, we used a bag-of-words and one-hot encoding to represent the events as well as constructed a state graph to use as a GNN embedding in the LSTM model."}]},"datasets":{"status":"available","items":[{"name":"semi‑synthetic dataset of customer interactions","provider":null,"kind":"synthetic","access":"unknown","period":null,"universe":null,"quote":"Our experimental setup consisted of the semi-synthetic dataset with 12 thousand customer interactions, split into 70% training, 15% validation, and 15% test sets."}],"known":[],"sample":{"start":null,"end":null,"market":null,"asset_class":null,"frequency":null},"data_resolution":"synthetic"},"transaction_costs":{"status":"not_applicable","modeled":"n/a","headline_net_of_costs":null,"detail":null,"quote":null,"source":"front_matter","rule":"R3","candidates":{"front_matter":"n/a","evidence_card":"n/a","evidence_card_quoted":false},"meaning":"the paper reports no trading or execution result"},"results":{"status":"available","evaluation_type":"forecast-accuracy","findings":[{"finding":"Goal prediction accuracy reaches 71% for the LSTM and 77% for the GNN+LSTM model.","kind":"predictive-accuracy","quote":"... Table 3: Accuracy for each approach when predicting agent type."},{"finding":"Customer type prediction accuracy is roughly 90% or higher for both models across income, fail behavior, and digital behavior categories.","kind":"predictive-accuracy","quote":"LSTM 63% 89% 97% GNN+LSTM 70% 90% 97% Table 4: Accuracy for each approach when predicting customer type."},{"finding":"Trajectory prediction accuracy declines with horizon length, but GNN+LSTM consistently outperforms LSTM (e.g., 67% vs 52% for next 1 event).","kind":"predictive-accuracy","quote":"LSTM 52% 40% 33% GNN+LSTM 67% 62% 49% Table 5: Accuracy for approach when predicting the next 1, 5, and 15 customer events."}],"reported_performance":null,"out_of_sample":{"method":"holdout","headline_out_of_sample":null}},"independent_evidence":{"status":"none_found","supports":[],"contradicts":[],"qualifies":[],"compared_with":34,"citations":{"cited_by_in_corpus":1,"by_independent_authors":1,"how_cited":{"used_as_baseline":0,"uses_method_or_data":0,"extends":0,"discusses_findings":0,"mentions":0,"unclassified":1},"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":"2026-11-14T18:09:03Z"},"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/"}