Step 2. Collect the Real Dataset The dataset should contain at least 100 observations. For example, if your project is house price in year 2012 at San Jose, you need to collect at least 100 houses' prices at San Jose in year 2012, and collect these houses' properties (input variables), such as number of bedrooms, number of bathrooms, etc. If your project is country's life expectancy in year 2010, you need to collect at least 100 countries' life expectancy in year 2010, and collect these countries' properties, such as GDP, HIV rate, unemployment rate, etc