A machine learning model is a system that learns patterns from example data and uses them to make predictions on new data, rather than following rules a person wrote by hand. In commercial real estate extraction, it classifies and labels content, such as tagging a rent roll row as occupied or vacant, or a lease clause as a renewal option.
How Does a Machine Learning Model Work?
A machine learning model works by adjusting internal parameters to fit labeled examples, then applying that fit to unseen inputs. In supervised learning, each training example has a known correct output, and the model learns a mapping from input features to that output. Once trained, the model predicts a label for new data it has never seen, per GeeksforGeeks' overview of supervised learning.
Building a model has two stages: construction and use. During construction the model is fit on a training set where the true answers are known. During use, the fitted model assigns labels to new records. Performance is judged on a held-out test set, so the score reflects data the model did not learn from.
Stage | What happens |
Features | The input signals, such as a rent roll row's numeric fields |
Label | The correct output, such as "occupied" or "vacant" |
Training | The model tunes parameters to fit the labeled examples |
Inference | The trained model predicts labels for new records |
Why a Machine Learning Model Matters
A machine learning model matters in commercial real estate because deal documents do not follow one fixed template. A rules engine breaks the moment a landlord reformats a rent roll. A model learns the pattern behind many formats, so it generalizes to layouts it has not seen, which is what makes automated extraction hold up across a portfolio.
Accuracy alone can mislead, so operators judge a model on precision and recall. Precision is the share of predicted positives that were correct; recall is the share of actual positives the model caught, per Built In. Accuracy is a poor guide when classes are imbalanced: a rent roll that is 95% occupied lets a lazy model score 95% accuracy by calling every row occupied while catching zero vacancies.
Example
A machine learning model classifies 1,000 rent roll rows as occupied or vacant, and the results form a confusion matrix. The model returns 380 true positives, 20 false positives, 40 false negatives, and 560 true negatives, where "positive" means vacant. From these four counts every standard metric follows.
Metric | Calculation | Result |
Accuracy | (380 + 560) / 1,000 | 94.0% |
Precision | 380 / (380 + 20) | 95.0% |
Recall | 380 / (380 + 40) | 90.5% |
F1 score | 2 x (0.950 x 0.905) / (0.950 + 0.905) | 92.7% |
Accuracy reads 94%, but the 40 false negatives matter: those are vacant units the model called occupied, which would overstate income if trusted blindly. Recall of 90.5% exposes that gap where accuracy hides it. This is why a reviewer routes low-confidence rows to a human rather than accepting the headline accuracy figure.
Variations and Edge Cases
Machine learning models come in families defined by what they learn and how. The type chosen shapes what an extraction pipeline can label and how much labeled data it needs.
Variant | Treatment |
Classification model | Predicts a category, such as occupied vs vacant |
Regression model | Predicts a number, such as estimated market rent |
Supervised model | Learns from labeled examples, the common extraction case |
Unsupervised model | Finds structure in unlabeled data, such as clustering |
Large language model | A text-generating model, one specialized subtype |
Machine Learning Model vs Rules-Based System
A machine learning model is often confused with a rules-based system, and the difference is where the logic comes from. A machine learning model learns its logic from labeled data and generalizes to new formats. A rules-based system follows conditions a developer wrote by hand and does exactly that, no more.
Rules are predictable and easy to audit but brittle: a new rent roll layout can break every rule at once. A model degrades more gracefully across unseen formats but needs training data and monitoring. Extraction pipelines often pair them, using rules for fixed fields and a model for the variable narrative in a lease.
Frequently Asked Questions
What is a machine learning model in simple terms?A machine learning model is a system that learns patterns from example data and predicts outcomes on new data, instead of following hand-written rules. In commercial real estate it labels content, such as marking a rent roll row occupied or vacant, or identifying a clause in a lease.
What is the difference between accuracy, precision, and recall?Accuracy is the share of all predictions that were correct. Precision is the share of predicted positives that were truly positive. Recall is the share of actual positives the model caught. On imbalanced data such as a mostly occupied rent roll, recall and precision reveal errors that accuracy hides.
Is a large language model a machine learning model?Yes. A large language model is one specialized type of machine learning model, trained on text to predict tokens. Every large language model is a machine learning model, but many machine learning models, such as small classifiers, are not large language models.
Related Terms
Large Language Model
Training Data
Confidence Score
Field Extraction
Structured Data Extraction