What is Time Series Analysis?
Time series analysis helps us understand and learn about the changes in our data over time. It involves collecting data points at consistent intervals over a set period of time, and using statistical methods to identify trends, patterns, and relationships in the data.
Time series analysis can also be used for forecasting future data based on historical trends and patterns as we will see later in this post.
Qlik Demo showing UK Crime trends: Explore demo here
Where can it be used?
Time series is used in a variety of industries, including finance, economics, retail, and meteorology, to help understand trends/patterns or make predictions and forecasts based on historical data, or simply to develop insights into how things change over time.
For instance Retailers can use time series data in a number of ways to support their decision making and improve business performance, including:
Sales forecasting to project sales trends in the future and predict demand, plan for inventory, and optimize supply chain to avoid overstocking or shortages.
Customer behavior analysis to understand how their customers’ habits change over time and identify spending habits and purchasing patterns which can help optimize marketing targeting efforts.
Pricing analysis to understand how prices change over time affect sales and customer behavior.
Inventory management to identify demand for specific products.
Ways to visualize Time Series data:
When it comes to looking at time-based data, we usually tend to use a Line Chart as it offers a simple and quick way to view changes related to a measure (Y-axis) over the course of time (X-axis).
Avg Temp over time comparison between 3 countries
Other charts however can help visualize the information better, here are some examples:
- A heatmap chart as seen on Github’s commit history shows a darker colored squares when more commits are made during a day and lighter color for less commits allowing us to easily notice the more productive days.
-A simple slope chart can be used when the time interval is limited and we want to see direct transitions over time, for instance comparing total sales data by industry between two years. Check out my other post to learn more about how this slope chart was created using Qlik's open source libraries.
Time Series Forecasting in Qlik Sense
Time series forecasting allows us to predict where a particular metric is going to be in the future based on historical data.
Qlik Sense makes it easy to do this on the Line chart.
Let’s build a chart with forecast that looks at average temperatures in the US (The QVF for this example is downloadable at the end of the post)
Create a filter pane and select United States to limit data to one country.
Create a Line Chart using:
Dimension: MonthYear
Measure: Sum(AverageTemperature)
Expand the Measure field and turn ON the “Forecast” setting
Move to the Dimension field and you can see that 2 new settings have appeared:
Forecast steps: the number of data points to be forecasted (we choose 12). Keep in mind that the further we go out in time, the broader the confidence becomes due to the increased uncertainty.
Step size: the granularity of the forecast steps (we choose Month to forecast for the upcoming year)
You can also change the method used for the forecast (SSA or OLS) and the appearance of the trend line
Confidence level describes the probabilistic range of the forecast and can be adjusted from 0 to 1. If 0.9 is selected, that means that 90% of the time, we will be within the blue range.
Using Insight Advisor:
Qlik has recently added the “Trend with Forecast” analysis type to Insight advisor. You can simply use the suggested measure and dimension and generate the chart:
To learn more about time series forecasting, check out this help article.
Thank you!
...View More
Kick off the New Year with Qlik Cloud, as I will explain exactly what makes Qlik Sense the next-generation data exploration and visualization software. Register here
It's #Fridaythe13th - so here are some quick fun facts about the iconic movie franchise that I discovered using #QlikSense, and some I did not know about until now, so spoilers ahead - enjoy!
Qlik-cli is a great tool to use for automating tasks or making scripting easier, in this blog post I will share 6 cool things you might not know about Qlik-cli.
The NL Insights object, found in the Qlik Dashboard bundle, allows app developers to add natural language generated insights to a sheet. The NL Insights object can be dragged to a sheet, just link other chart objects. Developers can use the NL Insights object to display an analytical summary of the data. This can accompany a single chart, a sheet, or the entire app. Once measures and/or dimensions are added to the object, narrative insights are automatically generated. The developer controls what data they would like to provide insights for based on their dimension and measure selections.The image below is the Dashboard bundle.Qlik Dashboard bundle Properties window > Data sectionAfter adding the master measure Sales $ (as seen in the image on the left), the insights below are generated. This is an example of using the NL Insights object to provide overall insights for an app. By default, all appropriate analysis types are used to generate the insights based on the dimensions and/or measures selected, but the developer can modify these if they choose. Up to three dimensions and three measures can be added to a NL Insights object. If the combination of dimensions and/or measures does not return any insights, a message will be displayed.In the properties window, the developer can deselect any of the analysis types that they would like to remove from the auto-generated list. For example, in the image below, the selected analysis types can be removed by clicking on the “x” image next to the analysis type.Properties window > Appearance section > Analysis typesAccording to Qlik Help, NL Insights can offer the following analysis types:Calculated measureRankingRanking (grouped)Breakdown (geospatial)BreakdownOverviewRelative importanceYear to dateTrend over timeComparisonCorrelationProcess control (means)Natural language insights can also be used to accompany a specific chart. In the example below, a NL Insights object has the same dimension and measure that the line chart has and is used to provide a narrative of the chart. This can be useful to highlight important information. In this example, the style is set to bullet points, but it can also be set to sentences. Developers also have the option to set the verbosity to full or brief to show all recommendations or just the top recommendations. This example has verbosity set to full. Like other objects, insights are updated when selections are made in the app.Users can benefit from NL Insights – a native capability that provides AI generated insights to users. Developers can give users a narrative summary of a chart, a sheet, or an app, to supplement visualizations in an app. A few things to note about the NL Insights objects – you can have up to three NL Insight objects on a sheet and if you add a new master item to your app, you will need to refresh the app before the new master items can be used in the NL Insights object. Try the NL Insights object out in your next app. To learn more, check out this SaaS in 60 video and Qlik Help.Thanks,Jennell
...View More
Qlik Sense Map charts are used to geographically display data related to countries, cities, states, regions, or particular geolocations (etc…). Maps offer different ways to present your data by first setting a base layer, then adding multiple layers to the map which are specific locations highlighted in multiple ways including Area, Points, Lines, Density, boundaries etc..You can add as many layers as you want. These layers are comprised of dimensions and measures that allow to efficiently present geographical distribution of values related to locations in order to display a data story.Types of layers:Point Layer: composed of bubbles or markers (you can choose a shape) positioned in specific coordinates (Lat / Long)Area Layer: geographic shapes sized and placed on specific areasLine Layer: connect two fields containing point data or a field containing geometries.Heatmap Layer: uses color gradient to show data density (intensity is greater at the center and declines towards the outer perimeter)Base Maps:Using different base maps can enhance the way data is displayed and aid in analysis. You can choose from:Default: Good for when using only 1 or 2 simple layersPale: great for when you add multiple layers with different colors which might be difficult to read if using the default baseDark: good for when you color layers with bright colors, this style can also be used when it matches the rest of your dark themed dashboard.Satellite: gives a more realistic look to your mapNone: in this case, you can add custom background layers (covered later in this post)Best Practices when designing Maps:Make your maps Simple to read:Be sure to use colors purposefully (stick to the default palette or use something like this https://jarrettmeyer.com/2018/08/07/viridis-color-palette)Use tooltips when necessary to provide more context when your layers are hovered.Change fonts and font colors when necessary, think about contrast between your layer coloring and the overlapping text font color.Think about the data you are visualizing:Continuous vs DiscreteIf visualizing two or more variable, it’s better to use color and sizeFor instance, in a point layer showing office locations, you can color the point based on number of employees and size the bubble based on the Sum of sales made.Advanced Uses:Multi-Layer Maps:When including multiple layers in map chart, it might become hard to interpret data. In that case, you can address this by controlling at what zoom levels different layers appear or have layers that appear only if other values in a drill-down dimensions are selected. This allows to create multiple levels of detail as you make selections and zoom in and out or locations of interest on the map.Zoom-Dependent Layer DisplayLet’s create a map that relies on Zoom to reveal different layers.We start by creating the first layer as an Area layer to represent Countries and we color using the Sum(Sales) aggregated measure.We choose the Pale base type in order for the data to stand out more.Next, we add two new Point LayersFirst, to represent cities. We choose to color by measure using the Sum(Sales) aggregation.Second, to represent Customers. We choose the shape of the point layer to be a Triangle, and we color by Sum(Sales)Lastly, we configure the Layers to only show at certain zoom levels. We do that by changing the Layer Display “Show in zoom levels” property from Auto to Custom.Country Area Layer will show up until 4x zoomCity Point Layer will show from 4x to 9x zoomCustomer Layer will show starting at 10x zoomThe result:Drill-Down Layer DisplayLet’s create a map that uses a drill-down dimension to display layers based on selection. Keep in mind that Drill-down dimensions should have the fields in order of highest geographical are to smallest geographical area.First, we create a Drill-down master dimension that contains three levels:Sales RegionStateSales officeNext, we add out 3 layers:An Area Layer for Sales Region: Choose the created Master Dimension as the dimension of the layer, and choose Sales Region.Area as the location.An Area Layer for the states.A Point layer for Sales offices.To enable the drill-down functionality. For each layer, navigate to Layer Options > Display. And change the “Visible drill-down levels” property from Auto to Custom and un-select the non-relevant layers, for instance:The result:Tip:If you load data and it appears incorrect like below:Head to Location, switch off Scope for location from Auto to Custom.Change Location Type to “Administrative Area (Level 1)” in our case. Then, change Country to your location, in our example it’s ‘US’Multiple Background Layers:To use custom base maps beyond the types mentioned above, you can add different background layers.Let’s create a map with multiple background layers using a TMS and two WMS.First, we need to set the Base Map to non under Map Settings.Next, we add the first layer as a Background Layer and choose Format to be TMSWe include the appropriate URL and attributionFor the second and third layers, we use WMS server URLs, set the version and Load the WMS.Important: Keep in mind that when using URLs for TMS and WMS formats for background layers, these URLs that contain resource requests to external resources must have its origins allowlisted in the Content Security Policy, else the resource will not be loaded. WMS resources must have both image-src and connect-src directives allowlisted. More Info here.Finally, we add a Point Layer.The Result:The QVFs for all three advanced examples can be found below. You can load them to your Qlik Cloud tenant, investigate the chart settings, and tweak the configurations to practice these concepts.
...View More