Is Data Important

Today, many systems may generate huge amounts of data such as system logs, financial transactions, customer profiles, security incidents, and so on. It is encouraged by the advancement of some technologies like IoT, mobile devices, and cloud computing. There are also fields that specifically learn to manage and process a lot of data like data science and machine learning.

A set of data can be processed to produce certain results like detecting anomalies, predicting the future, or describing the state of a system. To generate such a result, the typical phases are collecting data, data preparation, visualization, and data analysis or generating results.

In collecting data, we have to take some considerations including the location where the data will be stored, the type of stored data, and the retrieval method or how other systems can consume the data. When we want to select a location, we should consider whether the storage is available in the cloud or on-premise infrastructure, whether it will be deployed in a single instance or in a cluster, or whether it uses a document or relational database. After we decide on the location, we should think about how the data is stored, what is the form, and the data type. Data pipelines may become a topic in this step to tackle issues in scalability, data source integration, and automating the collecting process.

In data preparation, we may execute several tasks including tidying up data, removing duplication, correcting data types, and handling missing values. When we find out some missing values in a record, first we need to think about the possible cause of it, then we can choose between inputting appropriate values or completely dropping the record. The appropriate value can be the mean, median, or maximum/minimum value depending on the case.

Then, visualization is needed so that the prepared data can be easily understood by representing it in a suitable format or helping analysts gain insight or describing the condition of something. Things to be considered in preparing visualization like accessibility and readability.

The final phase is result generation which can be in various forms depending on the initial intention. We may perform simple analytical procedures or advance machine learning techniques to generate complex results such as to make predictions or object clustering. We may run an A/B test when we want to understand the impact of changes in certain aspects of a system. We may run a supervised machine-learning technique to make a prediction based on predefined labels and available features. When we are not sure what information can be retrieved from the set of data, an unsupervised machine learning technique may be performed to provide clustering of data so that we can be helped in making the conclusion.

Based on the phases explained above, there are several roles that focus on a specific phase in data processing. A data engineer focuses on creating a data pipeline and preparing data so that data can be stored and consumed by any parties in the process. A data analyst focus on creating the visualization and data preparation for describing the retrieved information. An analyst may utilize tools such as Power BI or spreadsheets. In gaining insight or making predictions, a data scientist comes in. Programming skills and knowledge of statistics are necessary in this case. When it comes to generating prediction, reasoning, or classification, a machine learning scientist is required.