Data Analyst Interview Questions and Answers

• Data Mining: Data mining does not require any hypothesis and depends on clean and well-documented data. Results of data mining are not always easy to interpret. Its algorithms automatically develop equations.
• Data Analysis: Whereas, Data analysis begins with a question or an assumption. Data analysis involves data cleaning. The work of the analysts is to interpret the results and convey the same to the stakeholders. Data analysts have to develop their equations based on the hypothesis.

• Data Exploration: After identifying the business problem, a data analyst has to go through the data provided by the client to analyze the cause of the problem.
• Data Preparation: This is the most crucial step of this process wherein any data anomalies (like missing values or detecting outliers with the data have to be modeled in the right direction.
• Data Modeling: This step begins once the data has been prepared. Modeling is an iterative process wherein a model is repeatedly run for improvements. This ensures that the best possible result is found for a given problem.
• Validation: In this step, validation of the data is done between the one provided by the client and the model developed by the data analyst. The aim is to find out if the developed model will meet the business requirements or not.
• Implementation of the Model and Tracking: This is the final step, where the model is implemented in production and is tested for accuracy and efficiency purpose.

Some of the best practices for data cleansing includes:

• Development of quality data plans to identify where maximum data quality errors occur, to assess the root cause and design the plan accordingly.
• To follow a standard process of verifying the critical data before the creation of a database.
• To identify any duplicates and validate the accuracy of the data to save time during analysis.
• Tracking all the cleaning operations performed on the data is essential to repeat or remove any operations as necessary.

• Problem definition
• Data exploration
• Data preparation
• Modeling
• Validation of data
• Implementation and tracking.

• Tableau
• Rapid Miner
• Open Refine
• KNIME
• Google Search Operators
• Solver
• Node XL
• IO
• Wolfram Alpha’s
• Google Fusion tables

• Common misspelling
• Duplication in entries
• Missing out the values
• Illegal values
• Varying value representations
• Identifying overlapping data

• Missing completely at random
• Missing at random
• Missing that depends on the missing value itself
• Missing that depends on an unobserved input variable

• Data screening
• Data verification

To deal with multi-source problems one should:

• Get involves in a restructuring of schemas, to accomplish schema integration.
• And, Identify similar records and merge them into a single document containing all relevant attributes without redundancy.

• Univariate
• Multivariate

• Preparation of a validation report that gives information on all suspected data. Information like validation criteria that it failed and the date and time of occurrence should be taken care of.
• Experience personnel should examine the suspicious data to determine their acceptability.
• Invalid data should be assigned and replaced with a validation code.
• To work on missing data best use of analysis strategy like deletion method, single imputation methods, model-based methods, etc. should be followed up.

• Initially there is a need to consult with the stakeholders of the business to understand the objective of classifying the data. Then, pull new data samples and modifying the model accordingly and evaluating it for accuracy. For this, a necessary process of mapping the data, creating an algorithm, mining the data and visualization is done. However, one can accomplish this in multiple segments by considering the feedback from stakeholders to ensure that model can produce actionable results.
• A model does not hold any value if it cannot produce actionable results, an experienced data analyst will have a different strategy based on the type of data being analyzed.

To say a model is good, the following points need to be considered.

• The developed model should have predictable performance.
• It should be adaptable easily to any changes as per business requirements.
• It should be scalable to any data change.
• A model should be efficiently consumed for actionable results.

Data Screening: In this step various algorithms are used to screen the entire data to find any erroneous or questionable values.

Data Verification: In this step each suspect value is evaluated on a case by case basis, and a decision is made if the values have to be accepted as valid, rejected as invalid or if they have to be replaced with some redundant values.

• Firstly, partition the data into three different sets Training, Testing and Validation.
• Secondly, show the results of the validation set to the business owner by eliminating biases from the first two sets. The input from the business owner or the client will give an idea of whether the model predicts customer churn with accuracy and provides desired results or not.
• Data analysts require inputs from the business owners and a collaborative environment to operationalize analytics. To create and deploy predictive models in production there should be an effective, efficient and repeatable process. Without taking feedback from the business owner, the model will be a one-and-done model.

Data Mining is the process to extract the data or required information from the databases. the data extracted can be used for finding solutions or in the implementation of data.