data-analysis

Read only a small sample first using pd.read_csv with nrows=5 to inspect columns and dtypes before writing full analysis code. For large files, use chunking or sampling rather than loading everything into memory at once.

Use dedicated document tools to read DOCX and PDF files, then extract the target data and save it as a CSV file. Once converted, the data can be loaded with Pandas and analyzed like any other tabular source.

Check completeness (nulls, empty strings, pseudo-missing values like -1 or 9999), uniqueness (duplicate rows and primary keys), accuracy (valid ranges and types), consistency (units and enum values), and referential integrity across related files.

Pandas and NumPy handle core processing, SciPy and scikit-learn cover statistics and mining, and Matplotlib, Seaborn, and Plotly produce static or interactive charts. The tabulate library formats Markdown tables for reports.

Encoding mismatches are the usual cause; retry reading with encoding='gbk' or encoding='latin1'. If all data appears in one column, the delimiter is wrong, so try sep=';' or sep='\t'.

Yes, reports can be delivered as Markdown, PDF, or DOCX. PDF output is typically produced via HTML conversion tools like WeasyPrint or Pandoc, while DOCX reports are built with python-docx using native headings, tables, and embedded images.