Normalization vs Standardization скачать в хорошем качестве

Normalization vs Standardization

Normalize vs standardize data in 100 seconds ⏱️ Data may vary in unit, magnitude and range. Some analytical methods are sensitive 😢 to this. Data or features must therefore be rescaled 📏. There are three commonly used feature scaling methods, what’s the difference? There are two mean reasons for data rescaling. Methods that compute the distance between the features are biased towards numerically larger values if the data is not scaled. Examples are k-means clustering and nearest neighbour classification. Another reason is that gradient descent converges much faster with feature scaling than without, e.g. regression and neural networks. CONVERT UNITS Sometimes we need to convert units, e.g. degrees Celsius to degrees Fahrenheit 🌡️ °C = (°F - 32) / 1.8 Feature scaling would help data to be in almost the same range: • decimal scaling • data normalization or rescales data to a standard 🔔 curve: • data standardization DECIMAL SCALING: moving decimal point. x’ = x / 10^j where j is the smallest integer such that max(|vi‘|) less than 1. x is the original value and x' is the scaled value. DATA NORMALIZATION: data is rescaled to range from 0 to 1. x’ = (x – min(x) / (max(x) – min(x)) x is the original value and x' is the scaled value. Normalization can be seen as a decimal scaling variant and is also known as Min-Max normalisation or Min-Max scaling. DATA STANDARDIZATION: data is rescaled to properties of a standard normal distribution with mean, μ=0 and standard deviation, σ=1. x’ = (x - μ)/σ x is the original value and x' is the scaled value. Standardization is also called, z-score normalization. The Big Question: normalize or standardize data? Normalization scales in a bounded range. This means that the impact of outliers is very large with normalization. Standardization is much less affected by outliers. Standardization assumes that your data has a Gaussian (🔔 curve) distribution, normalization has no distribution assumptions. Always rescaling? Data rescaling is not always necessary, for example tree-based methods are fairly insensitive to the scale of the features. The choice of the feature scaling method will depend on your problem and the analysis method. There is no hard rule to tell you when to normalize or standardize your data. TIMESTAMPS 00:00 Why rescaling? 00:49 Decimal scaling 01:00 Data normalization 01:12 Data standardization 01:25 Choice rescaling method VIDEOS IN THIS SERIES 🎞️ 📽️ Table vs Matrix    • Table vs Matrix: what are the differences?   📽️ Why Network Analysis:    • Видео   📽️ How Symbio6 Improves Your Business:    • Видео   Get started with network analysis today! It has never been so easy to improve your business, we are 😀 to help you. https://symbio6.nl/en/ ABOUT US Symbio6 gets the best out of relational data to help organizations perform even better. We do this by helping to discover, collect, store, analyse and visualize relational data and convert this into concrete actions to 📈 your business. https://symbio6.nl/en/about-us #DataPreparation #StartNetworkAnalysisToday #Symbio6