A Hybrid Quantitative Framework for the Chronological Reclassification of Indeterminate Ceramic Assemblages

The chronological classification of ceramic assemblages is frequently limited by the presence of large groups of fragments lacking secure phase attribution. These residual categories, often labelled as “indeterminate”, “no phase” or “SIN_INFO” (“no information”), distort phase distributions, weaken diversity and equitability indices, and reduce the interpretative coherence of archaeological datasets. This study develops a hybrid and fully traceable quantitative framework for the chronological reclassification of indeterminate ceramic materials, applied to the Neolithic assemblage from Tossal de les Basses / Cerro de las Balsas (Alicante, Spain). The initial dataset comprised 7,507 ceramic fragments and 387 stratigraphic units, with approximately 47% of fragments (NTE) and 52% of minimum number of individuals (MNI) lacking chronological attribution. The proposed protocol combines three complementary procedures: cosine similarity between techno-decorative profiles and phase centroids, supervised classification through Random Forest, and proportional/profile-based imputation. These outputs were integrated through a hierarchical decision model that prioritised agreement between methods, high-probability predictions and controlled similarity thresholds, while reserving lower-confidence profile-based assignment for unresolved cases. Rather than eliminating uncertainty, the framework transforms it into explicit, ranked and reproducible decision rules. The results enabled the formal reintegration of previously indeterminate materials into the chronological sequence of the site, improving stratigraphic consistency, phase comparability and the interpretative use of ceramic diversity indicators. The complete workflow was implemented through Python scripts, structured matrices and documented procedures, ensuring replicability, auditability and transferability to other archaeological contexts affected by high levels of ceramic indeterminacy陶瓷组合的编年分类常因大量缺乏可靠阶段断代的陶片群体而受限。这类常被标记为“未定”“无阶段”或“SIN_INFO（无信息）”的剩余类别，会扭曲阶段分布、削弱多样性与均匀度指数，并降低考古数据集的解释一致性。本研究针对不确定陶瓷材料的编年重分类问题，构建了一套可完全溯源的混合定量分析框架，并将其应用于西班牙阿利坎特省托萨尔德莱斯巴斯/塞罗德拉斯巴尔萨斯遗址的新石器时代陶瓷组合。初始数据集包含7507件陶片与387个地层单位，其中约47%的陶片（NTE）与52%的最小个体数（MNI）缺乏编年断代信息。 本研究提出的流程结合了三项互补方法：技术-装饰特征剖面与阶段质心的余弦相似度分析、基于随机森林（Random Forest）的监督分类，以及基于比例/特征剖面的插补法。通过层级决策模型整合上述三类输出结果，该模型优先考量不同方法间的一致性、高概率预测结果与可控的相似度阈值，并为未解决的案例保留基于特征剖面的低置信度分配方案。该框架并未消除不确定性，而是将其转化为明确、分级且可复现的决策规则。 研究结果实现了此前未定材料的正式归位，将其纳入遗址的编年序列，提升了地层一致性、阶段可比性以及陶瓷多样性指标的解释性应用。完整的分析流程通过Python脚本、结构化矩阵与标准化文档化流程实现，确保了可复现性、可审计性，以及可迁移至其他受高比例未定陶片影响的考古场景。