preprocessing

AI systems learn patterns and make outputs from the data supplied to them. If the data contain wrong values, missing information, or systematic bias, the system may learn an unsuitable pattern and produce inaccurate or unfair results. Incorrect data do not guarantee correctness or have no effect, and they do not normally make the machine shut down immediately.

Data cleaning prepares a dataset by finding errors, duplicates, missing values, inconsistent formats, and irrelevant records. Appropriate correction, removal, or documentation makes the remaining data more suitable for analysis and training. Option A is correct. Cleaning does not eliminate all data or allow prediction without data; it improves the quality of useful data.

Data cleaning identifies problems such as duplicate records, missing values, incorrect entries, inconsistent formats, and irrelevant information. Correcting or appropriately removing these problems gives the model more reliable input and can improve its performance. Cleaning does not mean hiding outputs or abandoning privacy protections.

An AI system learns relationships and patterns from the information it receives. Incorrect or missing data can cause it to learn distorted patterns, overlook important cases, or produce unreliable predictions and decisions. Good-quality, relevant, and sufficiently representative data generally supports more dependable results, although testing is still required.