# Fuzzy Logic and Predictive Analytics – Prof. Sriram Prabhakar

18 Aug 2023.

Fuzzy logic technology is a decision-based reasoning process that is almost identical to human thinking, in contrast to binary logic, which only works in a true or false sense. This approach incorporates all of the possibilities between yes and no and is similar to how people make decisions. Fuzzy logic is used in neural networks, control systems, and household appliances like rice cookers to make subtle adjustments based on a variety of criteria. This is particularly useful when dealing with various variables that cannot be categorised using a simple yes/no method.

The fuzzy sets theory was first introduced in 1965 by Professor Lotfi Zadeh of UC Berkeley. This established the foundation for fuzzy logic, an idea he also put forth in 1973. Mathematical sets, or collections of things called elements, are the focus of the idea of fuzzy sets.
In most mathematical sets, an element either belongs to the set or it does not. On the other hand, fuzzy logic allows for elements to have varying degrees of set membership. A chatbot learning human language is an example of how fuzzy logic is used in machine learning to help computers see the world more humanely and learn new things.

Control systems frequently include fuzzy logic because it may be used to simulate human decision-making. For instance, a robot arm could be controlled using a fuzzy logic controller. The controller would take into account the arm’s current position, the target position for the arm, and the arm’s movement speed. It would then determine the most effective way to move the arm into the required posture using these inputs.

Data mining also makes use of fuzzy logic. The technique of removing valuable information from huge data sets is known as data mining. Fuzzy logic can be used to identify patterns in data and anticipate what will happen in the future. A data mining algorithm might be used, for instance, to forecast whether a customer will purchase a particular item.

Artificial intelligence (AI) in the form of fuzzy logic is founded on the principle of approximative reasoning. This kind of AI is frequently employed in circumstances when it is necessary to make conclusions based on unreliable or partial data.

Over other AI techniques, fuzzy logic provides a lot of advantages. Its ability to handle erroneous or incomplete data is one of its advantages. This is due to fuzzy logic’s usage of a set of rules that are more focused on generalisations than on details. This means that even with inadequate data, it may still make conclusions that are reasonably correct.

Fuzzy logic also has the advantage of being more adaptable than other AI techniques. This is so that it can be readily altered to deal with new information or conditions that arise. Due to its adaptability, it is highly suited for applications where the data or circumstances are dynamic, such stock market forecasting or weather forecasting.

Finally, compared to other AI techniques, fuzzy logic is frequently easier for people to understand. This is because it employs a set of straightforward rules that are basic enough for humans to understand. As a result, it is simpler for humans to comprehend how the AI system decides and, if necessary, to fix and enhance the system. Fuzzy logic is founded on the notion that there are numerous viable solutions to a problem, each of which may be appropriate in a particular situation, rather than a single proper solution. This gives decision-making greater leeway and can provide better outcomes.

Fuzzy logic is a strong tool for AI applications because of a variety of benefits. It can handle erroneous data, is adaptable to change, and is easier for people to understand. It is a useful tool for resolving complicated problems because of these benefits.

Fuzzy analytics (FA), which is the use of fuzzy sets, fuzzy logic, and all related fuzzy methodologies in the field of marketing, has received more attention in recent years. Although it has been used in numerous marketing research, the notion is still not well understood, which hinders the theoretical and practical advancement of marketing science.

To boost performance, many businesses use Business Intelligence and sophisticated data analysis. The management of data from many data sources is a part of these analytical processes. When combining the datasets, there can be a disagreement between different naming conventions. Duplications and inaccuracies emerge from this, which leads to inconsistent reporting and false business results.
Think about a social media dataset’s region column. ‘IND’ is used to represent one of the values in this case, but ‘India’ is the preferred representation. On the other side, a comparable dataset substitutes “IN” for “India” instead. As a result, while combining these datasets, the aforementioned record won’t be linked to the region column.

Due to differing notations, the Business Name (account name/customer name) fields tend to have the most incompatibilities when connecting the data. These could be found in transactional, social media, marketing, and sales data.

External sources of data include data from social media and information from other vendors. Internal sources of data include sales, marketing, transactional, and other types of internal data.
As a result, if we attempt to combine all of these data sources, there may be several duplications as a result of inconsistent or multiple naming conventions. As a result, this might not correctly reflect the business analysis of the company and the recommendation that followed from this data.

Several naming conventions can have negative effects, such as duplicate entries, inaccurate insights, untagged accounts, and a lack of integrity.

Therefore, the optimum option would be to automatically combine related data using sophisticated text analytics algorithms.

Integration Solution

Using powerful text analytics along with well-liked algorithms like FUZZY LOGIC is a workable answer to the aforementioned problems.

Using fuzzy logic to standardise account names

In the example below, we’re attempting to combine three distinct data sets from three distinct sources with company account names in a variety of formats. However, by applying the Fuzzy Logic technique, we can combine the names of all three accounts into a single record.

As a result, the algorithm will group all versions of Motorola and assign them to a standardized notation of Motorola.

he following are some of fuzzy logic’s main benefits:

β’ One Single Source of Truth;

β’ Accurate insight and summary of the data;

β’ Eliminates duplicates;

β’ Automated process of identifying similarly sounding company names;

β’ Alternative text mining process as opposed to manual clean-ups which typically result in human error (Automation)

How is fuzzy logic implemented?

For each combination of account names from various sources, the text analytics system determines the word pairs. The computer searches for words with similar sounds based on several criteria, recognises those reports from different sources, and groups them together to produce a single name.
The method is then executed to produce the token sort ratio. This ratio assigns a percentage value and shows how far apart the words are. The fuzzy logic token sort ratio helps accurately identify the exact word distance between two values.

Token sort ratio:

75-100% – Exact or almost matching

60-75% – A fair game

40-60% – Minor match

In conclusion, normalising datasets with fuzzy logic algorithms helps address a variety of practical issues that organisations have with data cleansing. Additionally, this method gets rid of inconsistent word matches. Additionally, automated fuzzy algorithms reduce the amount of time needed for dataset integration and normalisation.