Similarity measure

What is meant by similarity measure in data mining?

The similarity measures in data mining, is a distance with dimensions describing object features. This means that in case the distance among two data points is small then there is a high degree of similarity among the objects and vice versa.

The similarity measure in data science is a way of measuring how data samples are related or close to each other. A dissimilarity measure is used to figure out how much the data objects are distinct.

These terms are generally used in clustering when similar data samples are grouped into one cluster. All other data samples are grouped into different ones. It is even used in classification, like k-nearest neighbors (KNN), in which the data objects are labeled on the basis of the features’ similarity.

The similarity is rather subjective and it has a heavy dependence on the context and application. As an example, the similarity among vegetables can be determined on the basis of their taste, size, color, etc.

Statistical measures of similarity enable scholars to think computationally about how alike or different their objects of study could be, and these measures happen to be the building blocks of many other clustering and classification techniques. In the context of text analysis, the similarity between two texts can be assessed in its most basic form by representing each text as a series of word counts and calculating distance using those word counts as features. 

The similarity measures in data mining is generally expressed as a numerical value - it increases when the data samples are more alike. It generally tends to be expressed as a number between zero and one by conversion: zero means that there is a low level of similarity(the data objects are dissimilar) and one means that there is a high level of similarity(the data objects are very similar).

‍

What is similarity?

Similarity is essentially a vast umbrella term that encompasses a wide range of scores and measures for assessing the differences among several kinds of data. Honestly, similarity refers to far more than you could really cover in one place.

‍

Why is similarity measure important?

Similarity measures are important because similarity is a vital element for understanding the ​​variables that motivate behavior and mediate affect.

Similarity also played a key role in psychological experiments and theories. As an example, in several experiments, people are asked to make direct or indirect judgments about the similarity of pairs of objects. Various experimental techniques are employed in these studies, but the most frequently used techniques involve asking subjects whether the objects are the same or different, or asking them to produce a number, between say 1 and 9, that reflects their feelings about how similar the objects appear (e.g., with 1 meaning very dissimilar and 9 meaning very similar).

The concept of similarity even plays an important (but less direct) part in the modeling of several other psychological tasks. This tends to be particularly true in theories of the recognition, identification, and categorization of objects, in which a common assumption is that the greater the similarity between a pair of objects, the more likely it is that one will be confused with the other. Similarity even happens to be very important in the modeling of preference and liking for products or brands, along with the motivations for product consumption. There is a common assumption here that when you’re evaluating a product, people tend to imagine an ideal and then judge the similarity of the offered product to this ideal.

‍

What is the difference between similarity and distance measures?

As mentioned earlier, similarity can refer to a wider category of similarity measures, but distance generally refers to a more narrow category that measures the difference in Cartesian space. 

In the context of text analysis, these concepts tend to be reciprocally related - distance is simply the opposite of similarity and vice versa. When you measure distance, the most closely related points will have the lowest distance, but when you measure similarity, the most closely related points will have the highest degree of similarity. 

‍

What are the types of similarity?

Here are some of the types of similarity:

Euclidean Distance

Euclidean distance is widely used as the traditional metric to work on problems with geometry. You can explain it as the ordinary distance between two points. It happens to be one of the most widely used algorithms in cluster analysis. If you’d like an example of an algorithm that makes use of this formula, just look at the K-Means clustering algorithm. Mathematically it will compute the root of squared differences between the coordinates between two objects.

Manhattan Distance (City Block Distance)

The Manhattan Distance or City Block Distance determines the absolute difference among the pair of the coordinates. It might be surprising, but the simplest way of calculating the distance between two points is to go horizontally and then vertically until you get from one point to the other, instead of just going in a straight line. It is a simpler technique since it only needs you to subtract instead of performing more complicated calculations.

Cosine similarity

Cosine similarity is quite different from the previous two. This similarity measure is more concerned with the orientation of the two points in space than it is with their exact distance from one another. It measures the level of similarity between two vectors of an inner product space. The cosine of the angle between two non-zero vectors is measured.

Jaccard similarity

The Jaccard distance measures the similarity of the two data set items as the intersection of those items divided by the union of the data items.

Minkowski distance

The Minkowski distance is the generalized form of the Euclidean and Manhattan Distance Measure.