Mobile phone data provide valuable insights into urban structure, yet traditional clustering methods assign a single label (e.g., home or work) per location, oversimplifying mixed-use areas. We present MAPLID, a supervised multi-label framework for place identification using Call Detail Records (CDRs) from Telecom Italia. The study area in Milan (23.5 km2) was divided into 10,000 grid cells, with a 20 × 20 subgrid (4.7 km2) manually labelled for home, work, and high/medium/low density. Several multi-label classifiers (Label Powerset, Binary Relevance, Classifier Chain, and ML-kNN) were evaluated using 10-fold cross-validation and metrics such as accuracy, F1-score, and Hamming loss. The best performance, achieving 88.3% average precision and leading in four of nine metrics, was achieved with Label Powerset and Random Forest applied to outgoing call data. MAPLID was further validated in the province of Trento, confirming its robustness in more irregular and forested terrain. Incorporating multimodal Point of Interest (POI) features from OpenStreetMap improved accuracy in data-sparse regions like Trento. The results demonstrate MAPLID’s capability to capture multiple concurrent urban functions and to generalise across contrasting geographic contexts, offering a computationally efficient framework for dynamic land-use mapping, urban planning, and mobility analysis using large-scale telecom data.