prophet-rb 0.1.0 → 0.1.1

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: b8cdf7fd6b309bc24d83bb12e4ad87ba1ba36ba0c6abc803485f11a5718ab649
-  data.tar.gz: d63c201452d57ef99349f3bf8b51249863cadbc555daef6a1f122d2ede571fdb
+  metadata.gz: e97133837196d4e1c97d69687d42f2d97e552d7be897f5e7a805efb5bab73e32
+  data.tar.gz: 4450d57d2c3da8632011f9f5a802891586b3e19347abe377e533ba5e8922708f
 SHA512:
-  metadata.gz: 81991c4edf0fa86d34e876ff8ebf2344491ae10dde95d736bbae3987592e7d419dc5167bf2b390eec19a88700152fff8c8f7eadb18f2d81fcfa928c9172c1ee8
-  data.tar.gz: 17e146ec6c6f74c792ac6da8bcd1103ad34df8303c7c76961ef956fcdbd0b343f14ff6a97ec4774c4d813dcfd07f5c3a1404365f24f27a636450f2bed9e895b0
+  metadata.gz: 6f0ed88d1a93d2f15e9750640833dbd889d8dea86255c8ec29c0fdc608ce27d17a0f617cbcaaee0be4b469b8e945f0ead9161875907a44a0555173e0f1a2c984
+  data.tar.gz: 485b4742b5267a8540445a87d59320a6ba5cc5589192369d22d69bfc1002d1ae2cb822a88a547ab63ff113e44b5ba47db51c45acb4bedc84079afd57210ea4ed

data/CHANGELOG.md

@@ -1,3 +1,10 @@
+## 0.1.1 (2020-04-10)
+
+- Added `add_changepoints_to_plot`
+- Fixed error with `changepoints` option
+- Fixed error with `mcmc_samples` option
+- Fixed error with additional regressors
+
 ## 0.1.0 (2020-04-09)
 
 - First release

data/README.md

@@ -10,6 +10,8 @@ Supports:
 
 And gracefully handles missing data
 
+[![Build Status](https://travis-ci.org/ankane/prophet.svg?branch=master)](https://travis-ci.org/ankane/prophet)
+
 ## Installation
 
 Add this line to your application’s Gemfile:
@@ -98,6 +100,45 @@ m.plot_components(forecast).savefig("components.png")
 
 ![Components](https://blazer.dokkuapp.com/assets/prophet/components-b9e31bfcf77e57bbd503c0bcff5e5544e66085b90709b06dd96c5f622a87d84f.png)
 
+## Saturating Forecasts
+
+[Explanation](https://facebook.github.io/prophet/docs/saturating_forecasts.html)
+
+Forecast logistic growth instead of linear
+
+```ruby
+df = Daru::DataFrame.from_csv("example_wp_log_R.csv")
+df["cap"] = 8.5
+m = Prophet.new(growth: "logistic")
+m.fit(df)
+future = m.make_future_dataframe(periods: 365)
+future["cap"] = 8.5
+forecast = m.predict(future)
+```
+
+## Trend Changepoints
+
+[Explanation](https://facebook.github.io/prophet/docs/trend_changepoints.html)
+
+Plot changepoints
+
+```ruby
+fig = m.plot(forecast)
+m.add_changepoints_to_plot(fig.gca, forecast)
+```
+
+Adjust trend flexibility
+
+```ruby
+m = Prophet.new(changepoint_prior_scale: 0.5)
+```
+
+Specify the location of changepoints
+
+```ruby
+m = Prophet.new(changepoints: ["2014-01-01"])
+```
+
 ## Holidays and Special Events
 
 [Explanation](https://facebook.github.io/prophet/docs/seasonality,_holiday_effects,_and_regressors.html)
@@ -141,7 +182,25 @@ Specify custom seasonalities
 m = Prophet.new(weekly_seasonality: false)
 m.add_seasonality(name: "monthly", period: 30.5, fourier_order: 5)
 forecast = m.fit(df).predict(future)
-m.plot_components(forecast).savefig("components.png")
+```
+
+Specify additional regressors
+
+```ruby
+nfl_sunday = lambda do |ds|
+  date = ds.respond_to?(:to_date) ? ds.to_date : Date.parse(ds)
+  date.wday == 0 && (date.month > 8 || date.month < 2) ? 1 : 0
+end
+
+df["nfl_sunday"] = df["ds"].map(&nfl_sunday)
+
+m = Prophet.new
+m.add_regressor("nfl_sunday")
+m.fit(df)
+
+future["nfl_sunday"] = future["ds"].map(&nfl_sunday)
+
+forecast = m.predict(future)
 ```
 
 ## Multiplicative Seasonality
@@ -156,6 +215,20 @@ future = m.make_future_dataframe(periods: 50, freq: "MS")
 forecast = m.predict(future)
 ```
 
+## Uncertainty Intervals
+
+Specify the width of uncertainty intervals (80% by default)
+
+```ruby
+Prophet.new(interval_width: 0.95)
+```
+
+Get uncertainty in seasonality
+
+```ruby
+Prophet.new(mcmc_samples: 300)
+```
+
 ## Non-Daily Data
 
 [Explanation](https://facebook.github.io/prophet/docs/non-daily_data.html)
@@ -166,8 +239,7 @@ Sub-daily data
 df = Daru::DataFrame.from_csv("example_yosemite_temps.csv")
 m = Prophet.new(changepoint_prior_scale: 0.01).fit(df)
 future = m.make_future_dataframe(periods: 300, freq: "H")
-fcst = m.predict(future)
-m.plot(fcst).savefig("forecast.png")
+forecast = m.predict(future)
 ```
 
 ## Resources

data/lib/prophet/forecaster.rb

@@ -156,7 +156,7 @@ module Prophet
           raise ArgumentError, "Regressor #{name.inspect} missing from dataframe"
         end
         df[name] = df[name].map(&:to_f)
-        if df[name].any?(&:nil)
+        if df[name].any?(&:nil?)
           raise ArgumentError, "Found NaN in column #{name.inspect}"
         end
       end
@@ -201,7 +201,7 @@ module Prophet
       end
 
       @extra_regressors.each do |name, props|
-        df[name] = ((df[name] - props["mu"]) / props["std"])
+        df[name] = ((df[name] - props[:mu]) / props[:std])
       end
 
       df
@@ -218,32 +218,40 @@ module Prophet
     end
 
     def set_changepoints
-      hist_size = (@history.shape[0] * @changepoint_range).floor
+      if @changepoints
+        if @changepoints.size > 0
+          too_low = @changepoints.min < @history["ds"].min
+          too_high = @changepoints.max > @history["ds"].max
+          if too_low || too_high
+            raise ArgumentError, "Changepoints must fall within training data."
+          end
+        end
+      else
+        hist_size = (@history.shape[0] * @changepoint_range).floor
 
-      if @n_changepoints + 1 > hist_size
-        @n_changepoints = hist_size - 1
-        logger.info "n_changepoints greater than number of observations. Using #{@n_changepoints}"
-      end
+        if @n_changepoints + 1 > hist_size
+          @n_changepoints = hist_size - 1
+          logger.info "n_changepoints greater than number of observations. Using #{@n_changepoints}"
+        end
 
-      if @n_changepoints > 0
-        step = (hist_size - 1) / @n_changepoints.to_f
-        cp_indexes = (@n_changepoints + 1).times.map { |i| (i * step).round }
-        @changepoints = @history["ds"][*cp_indexes][1..-1]
-      else
-        @changepoints = []
+        if @n_changepoints > 0
+          step = (hist_size - 1) / @n_changepoints.to_f
+          cp_indexes = (@n_changepoints + 1).times.map { |i| (i * step).round }
+          @changepoints = ensure_arr(@history["ds"][*cp_indexes].to_a.last(cp_indexes.size - 1))
+        else
+          @changepoints = []
+        end
       end
 
       if @changepoints.size > 0
-        @changepoints_t = Numo::NArray.asarray(((@changepoints - @start) / @t_scale.to_f).to_a).sort
+        @changepoints_t = (Numo::DFloat.cast(@changepoints.map(&:to_i).sort) - @start.to_i) / @t_scale.to_f
       else
         @changepoints_t = Numo::NArray.asarray([0])
       end
     end
 
     def fourier_series(dates, period, series_order)
-      start = Time.utc(1970).to_i
-      # uses to_datetime first so we get UTC
-      t = Numo::DFloat.asarray(dates.map { |v| v.to_i - start }) / (3600 * 24.0)
+      t = Numo::DFloat.asarray(dates.map(&:to_i)) / (3600 * 24.0)
 
       # no need for column_stack
       series_order.times.flat_map do |i|
@@ -848,8 +856,8 @@ module Prophet
     end
 
     def sample_predictive_trend(df, iteration)
-      k = @params["k"][iteration, true]
-      m = @params["m"][iteration, true]
+      k = @params["k"][iteration]
+      m = @params["m"][iteration]
       deltas = @params["delta"][iteration, true]
 
       t = Numo::NArray.asarray(df["t"].to_a)
@@ -907,12 +915,19 @@ module Prophet
     def make_future_dataframe(periods:, freq: "D", include_history: true)
       raise Error, "Model has not been fit" unless @history_dates
       last_date = @history_dates.max
+      # TODO add more freq
+      # https://pandas.pydata.org/pandas-docs/stable/user_guide/timeseries.html#timeseries-offset-aliases
       case freq
+      when "H"
+        hour = 3600
+        dates = (periods + 1).times.map { |i| last_date + i * hour }
       when "D"
         # days have constant length with UTC (no DST or leap seconds)
-        dates = (periods + 1).times.map { |i| last_date + i * 86400 }
-      when "H"
-        dates = (periods + 1).times.map { |i| last_date + i * 3600 }
+        day = 24 * 3600
+        dates = (periods + 1).times.map { |i| last_date + i * day }
+      when "W"
+        week = 7 * 24 * 3600
+        dates = (periods + 1).times.map { |i| last_date + i * week }
       when "MS"
         dates = [last_date]
         periods.times do
@@ -982,5 +997,10 @@ module Prophet
       u = Numo::DFloat.new(size).rand - 0.5
       loc - scale * u.sign * Numo::NMath.log(1 - 2 * u.abs)
     end
+
+    def ensure_arr(value)
+      value = [value] unless value.is_a?(Array)
+      value
+    end
   end
 end

data/lib/prophet/plot.rb

@@ -93,6 +93,24 @@ module Prophet
       fig
     end
 
+    # in Python, this is a separate method
+    def add_changepoints_to_plot(ax, fcst, threshold: 0.01, cp_color: "r", cp_linestyle: "--", trend: true)
+      artists = []
+      if trend
+        artists << ax.plot(to_pydatetime(fcst["ds"]), fcst["trend"].map(&:to_f), c: cp_color)
+      end
+      signif_changepoints =
+        if @changepoints.size > 0
+          (@params["delta"].mean(axis: 0, nan: true).abs >= threshold).mask(@changepoints)
+        else
+          []
+        end
+      to_pydatetime(signif_changepoints).each do |cp|
+        artists << ax.axvline(x: cp, c: cp_color, ls: cp_linestyle)
+      end
+      artists
+    end
+
     private
 
     def plot_forecast_component(fcst, name, ax: nil, uncertainty: true, plot_cap: false, figsize: [10, 6])
@@ -128,7 +146,7 @@ module Prophet
 
     def seasonality_plot_df(ds)
       df_dict = {"ds" => ds, "cap" => [1.0] * ds.size, "floor" => [0.0] * ds.size}
-      @extra_regressors.each do |name|
+      @extra_regressors.each_key do |name|
         df_dict[name] = [0.0] * ds.size
       end
       # Activate all conditional seasonality columns

data/lib/prophet/version.rb

@@ -1,3 +1,3 @@
 module Prophet
-  VERSION = "0.1.0"
+  VERSION = "0.1.1"
 end

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: prophet-rb
 version: !ruby/object:Gem::Version
-  version: 0.1.0
+  version: 0.1.1
 platform: ruby
 authors:
 - Andrew Kane
 autorequire: 
 bindir: bin
 cert_chain: []
-date: 2020-04-09 00:00:00.000000000 Z
+date: 2020-04-10 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: cmdstan
@@ -16,14 +16,14 @@ dependencies:
     requirements:
     - - ">="
       - !ruby/object:Gem::Version
-        version: '0'
+        version: 0.1.2
   type: :runtime
   prerelease: false
   version_requirements: !ruby/object:Gem::Requirement
     requirements:
     - - ">="
       - !ruby/object:Gem::Version
-        version: '0'
+        version: 0.1.2
 - !ruby/object:Gem::Dependency
   name: daru
   requirement: !ruby/object:Gem::Requirement