Você está se afogando em dados para suas visualizações. Como você decide quais pontos causarão maior impacto?

Identify the primary user and her/his primary question before even looking at the first data point. It's not the dearth of data that stops us from finding answers, it's the framing of right questions.

Ask yourself the business problems that needs solution? What questions are needing answers? The why is the reason behind the analysis and visualizations of data. When you identify the key problems needing answers in a range of complex dataset, it becomes easy to navigate in search of answers. This way your analysis is hinged on uncovering needed insights and so leaves your visualizations decluttered and with relevant details and information for enhanced data driven decision-making

Focus on the data points that align with the primary goal of the analysis. Remove the outliers and noise from that data points to bring consistency.

First of all focus on the outcome that we want to achieve after that data pre-processing then training model on the give dataset after that testing then deployment after that lastly , view the desired output.

D'après mon expérience l'identification des objectifs permet de plonger directement au cœur du projet d'analyse de données.la fixation des objectifs tiens compte du problème qu'on cherche à ressoudre et cela est fait avec les parties prenantes. En général il interviennent en amont du projet pour définir le problème, fixer les objectifs et définir les livrables avec le chef de projet et cela au travers des réunions de travail, afin de cerner clairement le besoin et ainsi circonscrire le travail a réaliser.

Anomaly Detection is the critical step after I am done with removing duplicates, validating data types, handling missing values, etc. Anomaly Detection or in simple words - finding unusual data points, might require domain-specific knowledge. But in many cases, it is as simple as common sense e.g. age > 120 for a person is definitely an inaccurate value. In other cases, you may utilize simple statistics to discover outliers. For example, measure Z-score. If it is too high, you are dealing with a potentially inaccurate data point. The threshold for Z-score may vary depending upon application but usually anything higher than 2 or 3 is a red flag.

Remove irrelevant or duplicate data. Correct inaccuracies and handle missing values to ensure data quality. Simplify datasets to highlight key insights.

In my experience it is also important to have a separation between generic and specific data cleaning processes. When its required each of these cleaning processes will come in picture as a robust structure for data cleaning - which not only would allow uniformity in data structures but also makes process streamlined for further maintenance.

The first step in the preparation for data visualization is to clean your datasets. A cleaned data set improves data reliability and credibility. To clean datasets is to remove inconsistencies, replace missing values, format accurately and remove duplicates. Without cleaning your datasets, you incur inconsistent visualization which is often misleading and inaccurate. A clean and error free dataset enables accurate and efficient visualizations leading to accurate and credible results for proper decision making

Here are the essential steps for data cleaning: 1. Remove Duplicates: Identify and delete duplicate records. 2. Handle Missing Values: Fill in missing values using methods like mean/median imputation, or remove records with missing data if appropriate. 3. Correct Errors: Identify and correct data entry errors, such as typos and incorrect values. 4. Normalize Data:Convert data to a common scale without distorting differences in ranges. 5. Standardize Formats:Ensure consistency in data formats (e.g., dates, currency).. 6. Validate Data: Cross-check the data against rules

Feature selection could be an iterative process, initially setting up reports with the most applicable feature and then iteratively changing the selection or customizing based on user feedback.

Select features that are in concordance with the variables you want to visualize. Rather than subjecting your visualizations with loads of irrelevant visuals. Identify the KPI's, trends, patterns and only choose variables that answer key questions and solve problems. This will afford you a precise yet effective data visualization. During the presentation of insights uncovered, your audience can grasp easily, the information delivered without struggling with loads of irrelevant visuals

Selecting features is generalizing the data as a whole. Like defining the class of a race in a DnD game, you will want to identify what key features make that data point as a whole. Following the DnD analogy, we want to choose what characterizes the data or 'character' (is it an elf class, does it have more stats in agility or strength? Does it only contain unique values? Is the character weak against ogres?). Those are the questions that should be asked when approaching what are the features of this.

Study your audience preference and style of communication during presentation. Use terminologies that your audience can grasp so that they understand the information you relay to them. Incorporate storytelling so that your audience can resonate with the facts you want to establish from the insights uncovered. Ask for feedback after each presentation in order to discover what you have done well and what needs to be improved on This way you constantly evolve and master the art of efficiently communicating data findings with your audience

Depending on the audience of anything related to public speaking or presenting, you will need to take note of who you will be presenting to (e.g. how do you approach a first grader and explain machine learning). Create the narrative that will drive the story to the point without referencing technical terms unless when necessary or if the audience is your peers. What do you want to have the audience take away from your story or findings.

Time to bust out the Crayola and begin building out what points you want to stand out. The point to detailing or visual encoding would be to relieve the audience having to look around to understand what they are seeing. Although it is possible to create an inspirational, monumental, amazing visual, you still are left with explaining about what the audience is looking at. when showing your findings, It would be good to highlight a datapoint/set that you would want to bring up in your presentation to segue into your findings (e.g. Following my research I have graphed what I have found and there is a correlation with xyz).

En este punto la simplicidad es nuestra aliada principal, hacer gráficos sencillos, fáciles de interpretar y libres de "ruido visual" es clave para la correcta interpretación de los datos por parte de los usuarios finales de las visualizaciones. El uso de los colores adecuados, respetando los códigos de color de las marcas y resaltando los insights o puntos claves de cada visualización ayudará a que las visualizaciones sean realmente útiles, amigables con los usuarios de los reportes. Acá se resalta también el conocimiento de la audiencia, pues se deben usar visualizaciones dependiendo de a quién vaya dirigido un reporte o informe.

El conocimiento del negocio es clave para las iteraciones de los reportes, ¿Qué tan dinámico es el negocio? ¿Cuándo es necesario realizar un ajuste al reporte? ¿Existen datos nuevos que ayuden a generar más insights? ¿Se están tomando decisiones basadas en los datos que entregamos desde las visualizaciones que construimos?. Estas preguntas nos ayudan a proyectar las iteraciones de los reportes y tener claro el rumbo de nuestros proyectos de visualización.

The value in being in all parts of this, allows you to update, rewrite, and scrap ideas that do not seem to go anywhere. Ask yourself if this has the possibility of supporting or opposing your original hypothesis. Revising your problem could be added on as a supplement that leads you to a greater discovery.

I would focus on relevance and clarity. I start by filtering the data to highlight key metrics aligned with my objectives. I use dimensionality reduction techniques like PCA to simplify complex data without losing essential information. I also employ aggregation to summarize trends and detect outliers to spotlight significant deviations. I prioritize data that tells a compelling story and supports my narrative, ensuring the visualization remains clear, insightful, and actionable.

Alles sehr schön. Aufgrund einer (persönlichen leidvollen) Erfahrung: mag, ich hier nochmal die Zielgruppe hervorheben, bzw. konkret das Management: Wenn sich das Management nicht für das Thema/Ziel interessiert, kann die Analyse und die Ergebnisse noch so spannend sein. Es wäre vergeblich.

Its always important to create virtual layers for your reports using mock or augmented data before connecting your actual data. Realizing the visualizations with augmented data would make them more robust and future proof, and could also allow your to foresee issues and ability to fix them.

Al iniciar un proyecto, construir un borrador que tenga las preguntas de negocio que se van a responder en cada parte de nuestra visualización, así evitaremos construir visualizaciones que no aporten valor a los usuarios finales.

It's always good to get a second pair of eyes on your visuals prior to presenting. Be prepared for questions if you take any.