Exploratory data analysis with CRISP-DM
Our standard way into any dataset. We understand the business first, then the data, then find what's worth acting on.
- Year
- 2024
- Tools
- R, Python, CRISP-DM, Data dictionaries
The aim
Use the CRISP-DM framework to run exploratory data analysis (EDA) on large datasets and get value out of them.
Why CRISP-DM
CRISP-DM (the Cross-Industry Standard Process for Data Mining) works in any industry, whatever the size or content of the data. It is iterative, so the pipeline keeps improving, and every step is checked against the project scope. It also puts business understanding first, so the analysis answers the question the business actually has.
What we did
- Defined the business objectives, then worked to understand the data.
- Prepared and cleaned the datasets. A data dictionary (built in R) records what each field means, where it comes from and how it is used.
- Used descriptive statistics, bivariate analysis and multivariate analysis to find trends, patterns and anomalies.
What we delivered
- An EDA report with the key findings.
- Visualisations of trends, correlations and outliers.
- Recommendations for using the insights in business strategy.
The outcome
Insights drawn from large datasets that lead to better decisions. Data quality and integration problems get caught before any model is built.