Applications of Data Science to business
By Angela C
September 20, 2021
Reading time: 5 minutes.
There are different types of data science including descriptive analytics, predictiive analytics and prescription analytics.
Descriptive Analytics can be used for business intelligence. To get useful data to the right people through the use of reports, dashboards etc. To identify trends in sales, customers that have churned, products that are performing well in certain regions, products with particular attributes that perform better than others etc.
It can be used to analyse customer behaviour, identify ways to make profit, compare data to competitors and to spot market trends. It can be used to track performance, to optimise operations, to predict success and identify potential issues.
Business intelligence (BI) combines business analytics, data mining, data visualization, data tools and infrastructure, and best practices to help organizations to make more data-driven decisions. 1
BI can help companies make better decisions using present and historical data within the business context. An analyst can use the business intelligence to provide perfomance and competitor benchmarks, allowing the organisation to run more smoothly and more competitively. BI can allow near real-time sales and performance tracking, providing the business with more insight into customer behaviour, identify market trends, forecast profits etc.
BI is a technology-driven process for analysing data and delivering actionable insights to help make informed decisions.
Whatever goals or questions the organisation has, data can be gathered and analysed to help it reach its goals.
Data gathered from the business activity of the organisation can be processed and stored in data warehouses. Data can be collected from internal IT systems or from external sources.
Data users can then access the data and analyse it to answer questions.
To run queries against the data, create data visualisations, dashboards and reports.
The analytics can be made more available to business users for strategic planning and decision making.
Data analysts and data scientists can use statistics and predictive analytics to discover patterns in the data and to forecast future patterns.
Business intelligence is used to answer specific questions and to help with decision making and planning. Asking questions can lead to futher questions and iteration.
The process is really a cycle of accessing data, making discoveries, exploring and sharing information.
Analytics allows business to react to evolving questions and demands. Business intelligence is now more interactive and available to more people and is no longer top-down coming in a static report from an IT department.
Many users across a department can customise their own dashboards and create reports much easier than ever before.
Data can be used to transform operations.
Data can be used to understand performance metrics, identify opportunities.
To identify if some regions are underperforming. To identify what is driving performance, to identify changes in clients requirements.
Business Intelligence can be presented through data visualisations. Data visualisations show data in a more understandable and accessible way. Dashboards can tell a story, highlight trends or patterns much more easily than just manually analysing the raw data.
IT departments generally manage the data in terms of security, accuracy and access while users can interact directly with the data.
Insights from artifical intelligence and machine learning can be incorporated into the business intelligence. Companies can become more data-driven.
Predictive Analytics uses machine learning to put data science models continually into production. Predictive analytics is used to predict what customers will churn, predict market performance, predict sales of a product in various locations, with diffferent features, under different market conditions etc.
Machine Learning is an application of artificial intelligence that builds algorithms and statistical models to train data to address specific questions without being explicitly programmed.
Prescriptive analytics or decision science uses data to help an organisation make decisions. To idenfity how to prevent particular types of customers from churning, how to market a product to optimise sales etc.
The Data Science Workflow
Data Collection: data is collected from various sources and stored for efficient access.
Data Exploration and Visualisation: Data is explored and visualised through the use of statistics, dashboards etc.
- Time-series tracks values over time;
- bar charts show comparisons across categories
- stacked bar charts tracks composition over time
Date Experimentation and Prediction: Using machine learning and artifical intelligence.
Dashboard tools include Excel, Google Sheets, Power BI, Tableau, Looker. Customised tools include Python plotly / dash, R shiny, D3.js etc.
Dashboards are useful when they will be used multiple times and when the information needs to be updated regularly. They are used when the requests are mostly the same all the time. However dashboards created using Plotly Dash allow users of the dahsboards to create their own customised dashboards.
Data Science Teams
A data team consists of various members with different skills:
- Data Engineers: Store and maintain data using SQL, Python, R, java etc.
- Data Analysts: Visualise and describe the data through the use of SQL, spreadsheets, tableau, python, R.
- Machine Learning Engineers: wrote code to predict with data using Python, R, Java etc.
- Data Scientists: build custom models to drive business decisions using SQL, Python, R etc.
Time-series forecasting predicts events through a sequence of time, capturing seasonality or periodic events.
Natural Language Processing (NLP) allows computers to process and analyse large anounts of natural language data. It takes text as input; performs word counts on important words in the text; word embeddings create features that group similar words
Deep Learning and Neural Networks enables unsupervised machine learning on data that is unstructured and unlabelled resulting in very accurate predictions.
Explainable AI is a set of tools and frameworks to help you understand and interpret predictions made by your machine learning models. With it, you can debug and improve model performance, and help others understand your models' behavior. 2